They show the destination, but they hide the road.

A strategy can finish with a beautiful equity curve and still spend years doing absolutely nothing. That dead period is called stagnation. It is one of the hardest parts of systematic trading because it messes with your head.

Before we go further, this article builds directly from two strategies I have already published. If you missed either of them, make sure to check them out first because today we are going to combine both into a portfolio and show why the pairing works so well.

Strategy 1: RSI2 ES.D + VIX Filter

Strategy 2: NQ ORB Strategy

Both strategies are strong on their own, but the real magic happens when we stop looking at them in isolation.

You can have a real edge, a profitable system, solid long term results, and still sit through months or years where nothing happens.

That is why I am always cautious when I see people promote one strategy like it is the answer to everything.

Especially the ORB.

Opening Range Breakout strategies are everywhere. YouTube, Twitter, Discord, paid groups, courses. Everyone has a version. Buy the high, sell the low, simple.

And to be fair, the idea has merit.

But long track records are rare for a reason. Most strategies have good periods and bad periods. A model can be hot for two years, then go flat for three. That does not mean it is broken. It means it is behaving like a strategy.

The real question is:

What do you do when your best strategy goes quiet?

That is where portfolio building comes in.

The Problem With Stagnation

Stagnation is the period between equity highs.

It is not always the same as drawdown. Sometimes the strategy is not losing much. It is just not making new highs.

That can be worse emotionally because you start asking the wrong questions.

Is it broken?
Should I stop trading it?
Should I increase size to make up for lost time?
Should I throw it away and start over?

This is why a single strategy is hard to trade.

Even a good one.

Below is our NQ ORB strategy. This is already a heavily improved version of the standard ORB you see promoted online. Our version has a much better structure, better average trade, fewer trades, and far less dead time than the original idea.

But even our version has long periods where performance is lacklustre.

Note: The pink shaded section highlights the stagnation period. This is the time where the strategy stops making new highs and effectively goes sideways. It may not look dramatic, but this is often the hardest part to sit through as a trader.

That is not a failure. That is reality.

Strategies breathe. They cycle. They go through periods where the market is not rewarding that behaviour.

The job is not to pretend stagnation does not exist.

The job is to build around it.

Strategy 1: Our NQ ORB Model

The first piece is our version of the Opening Range Breakout.

The classic ORB is probably one of the most promoted day trading strategies on the internet. The idea is simple, and that is why people love it.

Define the opening range.
Trade the breakout.
Close by the end of the day.

Simple. Visual. Easy to sell.

But simple does not always mean robust.

In the original ORB article, I showed why the classic version has problems when tested properly across a long sample. Our version rebuilt the idea with better structure, a daily filter, proper exits, and cleaner risk control.

Today, we are using that improved ORB as one half of the portfolio.

NQ ORB results

From the results:

• Total profit: $221,260

• Trades: 1,430

• Profit factor: 1.45

• Return/DD: 9.86

• Win rate: 46.15%

• Drawdown: $22,450

• Average trade: $154.73

This is a solid strategy.

But it still has periods where it goes quiet.

That is the point of today’s article.

We are not trying to pretend one good strategy is enough.

We are trying to build something more durable.

Strategy 2: ES RSI2 With VIX Filter

The second piece is our RSI2 strategy on ES with a VIX filter.

This is a completely different type of edge.

The ORB is an intraday breakout style model. It is looking for expansion after a defined opening structure.

The RSI2 strategy is different. It is a mean reversion style model. It looks for short term exhaustion, but only when the broader structure makes sense.

This is exactly what we want in a portfolio.

Different logic.
Different trade behaviour.
Different market exposure.
Different pain periods.

ES RSI2 results

From the results:

• Total profit: $352,800

• Trades: 177

• Profit factor: 3.28

• Return/DD: 14.17

• Win rate: 75.14%

• Drawdown: $24,900

• Average trade: $1,993.22

This is a very strong standalone system.

But again, even strong systems can go through quiet periods.

That is why we do not stop at one strategy.

The Superpower of Quant Trading

This is where quant trading has a real advantage.

A discretionary trader usually has one main style.

Maybe they trade breakouts.
Maybe they trade pullbacks.
Maybe they trade news.
Maybe they trade one market they know well.

That can work, but it creates dependence on one environment.

As systematic traders, we can do something different.

We can trade multiple edges at the same time.

Not just more trades.
Not just bigger size.
Actual different return streams.

This is the part I think most people miss.

The goal is not to take one strategy and size it until the account looks exciting.

The better path is usually to add exposure through different markets and different strategies, not simply by pushing position size higher on one idea.

That is a big difference.

If you just increase size on one system, you increase the same risk.

If you add another low correlation system, you may increase return while smoothing the ride.

This is why portfolio construction matters.

Ray Dalio has made this idea famous through his “holy grail” concept: build a group of good, uncorrelated return streams rather than relying on one source of return. We are not just stacking trades. We are stacking different behaviours.

Why These Two Fit Together

The ORB strategy and RSI2 strategy are almost perfectly uncorrelated in this test.

The correlation shown is:

• NQ ORB Strategy vs RSI2 ES.D + VIX: 0.02

That is exactly what we want.

A correlation of 0.02 means the two strategies are barely moving together.

That does not mean both make money every month.
It does not mean drawdowns disappear.
It does not mean the portfolio is risk free.

But it does mean one strategy is not simply a clone of the other.

That matters.

If ORB is quiet, RSI2 may still be working.
If RSI2 is flat, ORB may still be producing.
If both are working, the portfolio accelerates.
If one is in a bad patch, the other can help carry the load.

That is the entire point.

The Portfolio Result

This is where the benefit becomes obvious.

On their own, both strategies are strong. But the real improvement comes when we stop looking at them individually and start looking at them as a portfolio.

Return to Drawdown is one of the most important numbers here.

It tells us how much return we are getting for each dollar of drawdown. Higher is better because it means the strategy is using risk more efficiently.

The ORB alone had a Return / Drawdown of 9.86.

The RSI2 system alone had a Return / Drawdown of 14.17.

Combined, the portfolio moved to 18.34.

That is not a small improvement. That is the whole point of portfolio construction.

Compared to the ORB alone, the portfolio produced about 2.6 times the total profit, while drawdown only increased from $22,450 to $31,307.50. So yes, risk increased, but nowhere near the same rate as return.

That is what low correlation can do.

The portfolio did not just make more money because we blindly added more size. It made more money because the two systems behave differently. One is an ORB style strategy. The other is an RSI2 mean reversion style strategy with a VIX filter. They are not trying to make money from the same market behaviour.

And that is why the correlation of 0.02 is so important.

A 0.02 correlation means these two return streams are almost completely independent from each other. That is exactly what we want.

This is why I am a big believer in increasing exposure through more markets and more strategies, rather than just increasing position size on one system.

If you simply size up one strategy, you are increasing the same risk.

If you add another low correlation edge, you may increase return without adding the same level of pain.

That is the difference.

Stagnation Improves Too

Now let’s talk about the part most people avoid showing.

Stagnation.

The pink shaded section on the equity curves highlights the stagnation period. This is the time where the strategy stops making new highs and effectively goes sideways. It may not look dramatic, but this is often the hardest part to sit through as a trader.

The NQ ORB alone had a max stagnation of 2,521 days.

That is a long time to sit in a system waiting for it to make new highs.

Now look at the combined portfolio.

The max stagnation dropped to 1,017 days.

Now, you are probably thinking:

“Hold on, 1,017 days is still a long time.”

And you would be right.

This is not me pretending the portfolio is perfect. A 1,017 day stagnation period is still difficult to sit through. But the important point is the direction of improvement.

We went from 2,521 days of stagnation on the ORB alone to 1,017 days in the two strategy portfolio.

That is a reduction of roughly 60% from adding just one low correlation edge.

This is one of the clearest examples of why portfolios matter.

The ORB did not suddenly become perfect. The RSI2 strategy did not suddenly remove all pain. But when combined, the quiet periods became much more manageable.

Now imagine what happens as the portfolio grows.

If we continue adding robust strategies that are genuinely uncorrelated, across different markets, timeframes, and trade behaviours, we should keep reducing the reliance on any single edge.

That does not mean stagnation disappears completely. It does not mean drawdowns vanish. It does not mean the portfolio becomes risk free.

But it does mean the portfolio becomes less dependent on one strategy having a good period.

That matters in the real world.

Because the hardest part of trading a system is not usually the entry logic. It is sticking with it when nothing is happening.

Most people quit during stagnation.

They stop trading right before the next good period. Or they start changing rules. Or they add size at the wrong time. Or they abandon a real edge because it feels broken.

A portfolio helps reduce that pressure.

Not because it removes drawdown.

Not because it eliminates bad periods.

But because different systems can carry the load at different times.

That is the actual edge.

Not one perfect strategy.

A group of good strategies that do not all suffer at the same time.

Why Influencers Avoid Long Track Records

This is also why you do not see many long form, 20 year strategy tests from the average trading influencer.

Of course, some people do show real research.

But the vast majority do not.

Why?

Because long track records expose everything.

They expose stagnation.
They expose regime changes.
They expose drawdown clusters.
They expose whether the idea only worked during one easy period.

A strategy that looks amazing over six months can look very ordinary over 20 years.

That does not mean it is useless. It means you need to understand where it fits.

This is why I care less about a single perfect equity curve and more about how strategies behave together.

The Bigger Lesson

The big lesson here is not that ORB is good.

It is not that RSI2 is good.

The lesson is that edges have seasons.

Every strategy has a period where it looks brilliant.
Every strategy has a period where it looks broken.
The portfolio is what helps you survive the gap between those periods.

This is why I would rather build a book of strategies than keep increasing position size on one system.

More size on one edge means more exposure to one pain cycle.

More edges with low correlation means better diversification.

That is the game.

Final Thoughts

The NQ ORB model is a strong strategy.

The ES RSI2 with VIX filter is a strong strategy.

But together, they are more useful than either one on its own.

The portfolio has:

• Higher total profit

• Higher profit factor

• Higher Return/DD

• Better average trade than ORB alone

• Low correlation between components

• Improved stagnation profile

That is why I keep saying the same thing:

The goal is not to find one perfect strategy.

The goal is to build a portfolio of good strategies that behave differently.

That is how you create something you can actually stick with.

Disclaimer

This publication is for education and research only. It is not financial advice. Results shown are hypothetical backtests and simulations. Past performance does not guarantee future results. Futures and derivatives involve substantial risk. You can lose more than your initial investment. Always validate on your own data, costs, execution, and infrastructure before trading live.

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

💬 Message to readers

This week’s article is a continuation of our Turnaround Tuesday work.

Previously, we looked at Turnaround Tuesday as a swing trading idea using daily bars.

Turnaround Tuesday swing strategy article:

That article showed the core idea clearly. When the stock market starts the week weak, Tuesday often has a tendency to reverse higher. This is one of those simple seasonal effects that has been spoken about for decades because it actually makes logical sense.

Markets often overreact into the start of the week. Bad news gets priced in. Traders come in defensive. Selling pressure builds. Then, once the panic settles, the market can snap back.

That is the Turnaround Tuesday idea.

But this time, we wanted to test it differently.

Instead of holding a swing trade on daily bars, we built a day trading version on ES 15 minute bars.

Same broad market tendency. Different execution style.

The aim was simple. Can we use the daily Turnaround Tuesday edge as the setup, but execute it intraday so we are flat by the close?

That matters because not everyone wants overnight risk. Some traders prefer day session systems. Some traders want a cleaner margin profile. Some traders just want exposure to the edge without carrying the position into the next session.

This model is designed for that.

It uses a daily Turnaround Tuesday style condition to tell us when the market is weak enough to pay attention. Then it uses a 15 minute entry structure to avoid just buying blindly. Once in the trade, risk is controlled with an ATR stop. Everything is closed by 15:00.

Simple idea. Existing edge. Intraday execution.

That is what I like about it.

📖 The Idea

The idea starts with the classic Turnaround Tuesday tendency.

Stock indices often have a habit of bouncing after early week weakness. It does not happen every time, and it is not magic, but the tendency is strong enough that it is worth building around.

Most traders look at a weak start to the week and assume the market is broken.

Systematic traders should ask a better question.

Is this weakness setting up a short term rebound?

That is what this model is trying to capture.

It does not trade every Tuesday. It waits for a specific daily weakness condition first. Once that condition is present, it looks for an intraday long entry.

We also include Wednesday because the Turnaround Tuesday effect does not always resolve neatly on Tuesday itself. Sometimes the weakness carries into Tuesday and the rebound does not properly appear until Wednesday. By allowing Tuesday and Wednesday, the model gives the edge a little more room to express itself without turning the strategy into a broad day of week system.

The entry is also important.

We do not just buy at market because the setup is active. That would be too blunt. If the market is weak enough to trigger the setup, there is no guarantee it is ready to bounce immediately.

Instead, the strategy uses a stop entry above a calculated intraday reference level. This means price has to lift enough to trigger the trade. In plain English, we are not just catching weakness blindly. We are waiting for the market to show some intraday strength before getting involved.

The daily condition gives the model the reason to be interested.

The stop entry gives the model a cleaner way to participate.

The ATR stop controls the risk.

The end of day exit keeps the strategy clean and avoids overnight exposure.

That is the entire concept.

A day trading version of Turnaround Tuesday.

🧠 Why this works for ES

ES is one of the best markets for this type of idea because it has a strong tendency to mean revert after short term stress.

The S&P 500 is not like a random commodity contract. It has a long term upward drift, deep liquidity, and constant institutional participation. When selling pressure gets too aggressive over a short window, buyers often step back in quickly.

That is why Turnaround Tuesday has always been interesting.

It is not just a day of week trick. It is really a behaviour pattern.

Weakness into the start of the week can create forced selling, defensive positioning, and emotional exits. Once that selling slows, the market often has room to recover.

The day trading version tries to isolate that effect without holding overnight.

Instead of saying, “Tuesday is bullish, just buy,” this model says:

Is the market weak enough to justify the setup?

If yes, can we wait for intraday strength before getting involved?

That is a much better way to trade the idea.

The stop entry matters because weak markets can keep falling. A market order would enter immediately just because the setup exists. That is not what we want here. We want the setup first, then a small sign that buyers are actually stepping in.

That is why the entry is staged above an intraday reference level. If price never lifts enough, there is no trade. If price does lift enough, the model enters with a more structured trigger.

It still keeps the idea simple, but it avoids blindly buying weakness.

🎯 Why I like it in a portfolio

I like this one because it gives the portfolio a different type of return stream.

It is not a breakout model.

It is not a trend following model.

It is not a short volatility model.

It is a long only intraday seasonal mean reversion model built around a known stock market tendency.

That can be useful.

If you already run breakout systems, trend systems, bond systems, or commodity models, this brings something different to the table. It is trying to capture a specific index behaviour rather than just chase momentum.

The other reason I like it is the simplicity.

One broad setup. One entry structure. One stop. Flat by the close.

There is not much to hide behind.

And in my experience, the simpler the strategy, the easier it is to stress test properly and the easier it is to keep trading when it goes through a flat patch.

📈 Results of strategy since 1997

This strategy has been tested on ES 15 minute bars from 1997 to 2026.

The results are strong for a simple long only day trading model.

Net profit: $80,912.50
Profit factor: 1.52
Trades: 741
Win rate: 51.96%
Average trade: $109.19
Max closed trade drawdown: $8,237.50
Return to max closed trade drawdown: 982.25%
Trading period: 28 years, 8 months
Time in market: 1.02%

The part I like most is the sample quality.

This is not 80 trades over a hot recent period. This is 741 trades across multiple regimes, including the dot com period, the GFC, the low volatility grind, COVID, inflation, rate hikes, and the recent AI driven market.

That does not guarantee anything going forward, but it gives us a lot more information than a short recent backtest.

Equity of strategy

The equity curve is not perfect, and that is normal. No real strategy is. But the long term shape is what matters. The model keeps making new highs over a very long sample, while staying out of the market most of the time.

That is a good profile for a portfolio component.

🧪 Robustness testing

I trust a strategy only after it survives the usual stress tests.

Monte Carlo OHLC

Here we disturb the open, high, low, and close structure of the bars.

This is a useful test for an intraday model because small changes in bar structure can expose whether the strategy is too dependent on perfect historical prices.

If the strategy only works on the exact historical OHLC sequence, that is a warning sign.

Monte Carlo shuffle trades

Here we randomise the trade sequence.

This helps answer a simple question.

Did the strategy only look comfortable because the wins and losses arrived in a lucky order?

If the equity distribution still looks acceptable after the trade order is shuffled, that gives more confidence in the return profile.

Monte Carlo skip trades

Here we randomly remove trades from the sequence.

This is important because live trading is never perfect. You can miss trades, have execution issues, or have a data problem.

If missing some trades completely destroys the strategy, that is a problem. If the model still holds together, that is a better sign.

System Parameter Permutation test

We run many nearby combinations around the chosen settings.

If most of that neighbourhood still works, the chosen parameters are probably not just lucky.

This is especially important with simple systems. I do not want one perfect setting surrounded by failure. I want a broader area where the idea still behaves reasonably well.

Alternative timeframe test

We also tested the idea on nearby ES timeframes.

The goal here is not to find the best possible chart. The goal is to see whether the idea collapses the moment we move away from the original 15 minute timeframe.

If the strategy only works on one exact timeframe, that is a warning sign. If nearby timeframes still show similar behaviour, that supports the idea that the edge is broader than one chart setting.

Alternative market test: NQ

We apply the same idea to NQ without redesigning it.

NQ is more volatile than ES, so I do not expect it to behave exactly the same. But if the same broad logic still produces a reasonable equity curve, that tells us the tendency is not isolated to one symbol.

Alternative market test: YM

We apply the same idea to YM without redesigning it.

Again, the goal is not perfection. The goal is to see whether the Turnaround Tuesday concept has broader index behaviour behind it.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

💬 Message to readers

Today we start with one of the most promoted day trading ideas on the internet.

The Opening Range Breakout.

A quick search and you will find endless traders selling this as some kind of holy grail. Draw the opening range. Buy the high. Sell the low. Done.

Sounds easy.

The problem is that most of what you see online is not real validation. It is usually a few cherry picked charts, a short recent sample, or a discretionary version where the rules change depending on the day.

That is the problem with discretionary trading. A setup can look great right now, but how do you know if it worked through different regimes? How do you know if it survived 2008, 2020, low volatility, high volatility, trend, chop, and everything in between?

More importantly, how do you know when to stop trading it?

That is what we wanted to test.

First, we start with the classic ORB rules. No tricks. No extra filters. Just the basic version most traders talk about.

Then we rebuild it into something more robust.

The goal was not to create another complicated system. The goal was to take a popular idea and test whether it could be made genuinely usable with clean structure, a daily filter, intraday confirmation, and proper exits.

📖 What is the ORB strategy?

ORB stands for Opening Range Breakout.

The idea is simple. You define a range during the first part of the session. For example, from 8:30 to 9:00. That gives you the early high and early low of the day.

Once that opening range is complete, the classic rules are simple.

If price breaks above the opening range high, you go long.

If price breaks below the opening range low, you go short.

The theory is that the opening period acts like price discovery. Overnight orders, institutions, algos, retail traders, and macro news all collide during that early window. Once the market breaks away from that range, it can sometimes continue in that direction.

That is why the strategy is so popular.

It is simple. It is visual. It makes intuitive sense.

But intuitive does not mean robust.

So before building our version, we tested the classic version first.

🧪 Baseline: The classic ORB

The classic version uses the opening range high and low from 8:30 to 9:00.

After the range is complete:

Long entry is a buy stop at the opening range high.

Short entry is a sell stop at the opening range low.

The stop loss is placed at the opposite side of the range.

Everything is closed by the end of the day.

That is the classic ORB model in its purest form. The baseline code tracks the highest high and lowest low between the opening range start and end time, waits until the range is complete, then places stop orders at the ORB high and ORB low with stops on the opposite side of the range.

Classic ORB code

// Classic Opening Range Breakout

Inputs:
    ORBStartTime(830),
    ORBEndTime(900),
    DayExitTime(1500),
    MaxTradesPerDay(1);

// 1. Build the opening range between 8:30 and 9:00
ORBHigh = Highest high during opening range;
ORBLow  = Lowest low during opening range;

// 2. After 9:00, place breakout orders
If Time > ORBEndTime and MarketPosition = 0 then begin
    Buy next bar at ORBHigh stop;
    SellShort next bar at ORBLow stop;
end;

// 3. Use opposite side of range as stop
If MarketPosition = 1 then
    Sell next bar at ORBLow stop;

If MarketPosition = -1 then
    BuyToCover next bar at ORBHigh stop;

// 4. Flat at end of day
If Time >= DayExitTime then begin
    Sell this bar on close;
    BuyToCover this bar on close;
end;

That is the idea everyone knows.

Now here is the important part.

When you test the classic ORB over a long sample, it is not the clean money printer people make it out to be.

Original ORB

The original version did make money over the full sample, but the quality of the return was not good enough for me.

It produced a large number of trades, but the edge was thin.

The classic version produced:

Net profit: $24,930

Profit factor: 1.07

Trades: 6,502

Win rate: 41.80%

Average trade: $24.93

Max closed-trade drawdown: $83,613

Return to max closed-trade drawdown: 193.87%

That is the exact reason I do not like blindly trusting popular trading ideas.

The concept makes sense. But the raw version is too loose.

So we rebuilt it.

📖 The Idea

The improved version keeps the soul of the ORB, but it changes the way we use it.

The classic version treats every opening range as a fresh breakout level for that same day. Our version is more selective. It first asks the market to pass a daily filter, then uses the prior completed opening range level as the trigger area.

That matters.

The daily filter tells us whether the broader setup is worth paying attention to. The prior opening range level then acts as a reference point from the previous price discovery window.

In simple terms, we are asking two questions:

Is there a reason to be interested today?

And has price done enough around a known opening range level to justify a trade?

Sometimes price is already beyond that level when the order becomes active. In those cases, the entry behaves more like a market style execution because the level has already been cleared.

Other times, price has not cleared the level yet. In those cases, the stop order still matters because the market has to break through that prior opening range level before a position is opened.

That is the important distinction.

This is not just blindly buying every opening range high or selling every opening range low. It is also not purely a market order system. It is a daily filter plus prior ORB reference model. When price has already cleared the reference, the trade can trigger quickly. When it has not, the market still needs to prove itself by trading through the level.

That is why the stop-order entry still makes sense.

The ORB high or low represents the boundary of an early price discovery window. A stop order around that level means we are using that boundary as a trigger. If price is already beyond it, the market has already confirmed that level. If price is not beyond it, the market still has to break it before we take the trade.

Either way, the level is not random. It is anchored to the opening range, which is one of the most watched intraday reference points in index futures.

That is the difference between a social media ORB setup and a tested ORB model.

🎯 Why I like it in a portfolio

I like this model because it gives a portfolio something specific.

It is not a slow trend following model. It is not a mean reversion model. It is not trying to pick tops and bottoms.

It is an intraday momentum continuation model.

That means it can behave differently from systems that trade overnight, systems that fade extremes, or systems that rely on longer holding periods.

It also has a practical structure. It is flat by the close, uses a maximum of one trade per day, and only trades when the market gives both a daily and intraday reason to be involved.

That makes it easier to understand and easier to manage.

📈 Results of strategy since 2001

The improved version changed the profile dramatically.

Equity of improved strategy

Equity of improved strategy

The improved version produced:

Net profit: $192,390

Profit factor: 1.39

Trades: 1,425

Win rate: 45.61%

Average trade: $135.01

Max closed-trade drawdown: $22,744

Return to max closed-trade drawdown: 845.89%

Trading period: 25 years, 3 months

Time in market: 2.91%

Compared with the original ORB baseline, the improvement is substantial.

The baseline made $24,930 net profit.

The improved version made $192,390 net profit.

That is a 672% increase in net profit.

But the bigger story is not just the net profit.

The baseline version took 6,502 trades with a very small average trade and a large drawdown.

The improved version took only 1,425 trades, increased the average trade to $135.01, improved the profit factor to 1.39, and reduced the max closed-trade drawdown to $22,744.

That is exactly what a good filter should do.

Not make the strategy look pretty.

Make the strategy more tradable.

🧪 Robustness testing

I trust a strategy only after it survives the usual stress tests. (These tests only show the development window and not True OOS)

Monte Carlo OHLC

Here we disturb the open, high, low, and close structure of the bars.

This is a hard test for an ORB strategy because the model depends on intraday range structure. If a small change to the OHLC data destroys the edge, then the strategy is probably too fragile.

System Parameter Permutation test

We run many nearby combinations around the chosen settings.

If most of that neighbourhood still works, the chosen parameters are probably not just lucky.

For this test, the result is useful because the selected strategy sits inside a broader working area. I do not want to see one tiny island of profitability surrounded by failure. I want to see a neighbourhood where nearby settings still behave reasonably well.

Alternative timeframe test

We also tested the logic on alternative NQ timeframes.

The goal here is not to find a better curve. The goal is to see whether the idea collapses the moment you move away from the original chart.

If it only works on one exact timeframe, that is a warning sign. If nearby timeframes still show similar behaviour, the concept is more likely to be real.

Alternative market test: ES

We apply the same idea to ES without redesigning it.

If the results are still acceptable, that tells us the concept is not just a Nasdaq-only artefact.

Alternative market test: YM

We apply the same idea to YM without redesigning it.

Again, we are not expecting perfection. We are looking for evidence that the logic can travel.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

Larry Williams is known for simple ideas that cut straight to market behaviour.

No huge indicator stack.

No prediction engine.

No complicated regime model.

Just price, pressure, and trader behaviour.

This test started with one of those simple ideas: a Williams %R mean reversion strategy.

At its core, the strategy asks a very direct question:

Has the market become short-term oversold enough that a snap-back is likely?

Markets do not move in straight lines. Even during long-term uptrends, short-term selling can become emotional, stretched, and overdone. When that happens, mean reversion systems try to capture the rebound.

The original Larry Williams %R strategy did work.

But it had a problem.

It made money, but it was exposed to some ugly trade-level drawdowns when the market environment became unstable.

That is where the research became interesting.

The goal was not to make the strategy more complex.

The goal was to make it more selective.

What Williams %R Measures

Williams %R is a momentum oscillator created by Larry Williams.

It measures where the current close sits relative to the recent high-low range.

When %R is near 0, price is closing near the top of the recent range.

When %R is near -100, price is closing near the bottom of the recent range.

For mean reversion, the interesting area is the lower extreme.

A deeply negative %R reading tells us the market has closed near the bottom of its short-term range.

That does not guarantee a bounce.

Nothing does.

But in equity index futures, short-term downside pressure often becomes overdone. Sellers push too far, weak hands exit late, and the next move can become a relief bounce.

That is the behavioural foundation.

Williams %R identifies short-term downside exhaustion.

The strategy then waits for the market to recover enough to exit.

The original uploaded strategy is long only on ES.D RTH daily bars. It enters on an oversold Williams %R reading and exits when the close exceeds the prior bar’s high.

The Original ES Test

The original ES version produced a strong headline result.

On ES.D daily bars, single contract, no compounding:

This is a good strategy.

The win rate is strong.

The profit factor is strong.

The average trade is large enough to be meaningful.

But the equity curve shows the weakness.

The strategy had a large open trade drawdown event near the end of the test. That does not destroy the strategy, but it changes how I look at it.

A strategy can have great long term numbers and still be hard to trade if the worst trade gets too large.

That is exactly what stood out here.

The original strategy made money, but it was willing to buy oversold conditions even when volatility was already expanding.

That is dangerous.

Oversold in a normal market is one thing.

Oversold during a volatility spike is something else.

The first can mean exhaustion.

The second can mean the start of a deeper liquidation.

That is why I tested a volatility filter.

The Problem With the Original

The original Williams %R model treats all oversold readings the same.

But all oversold readings are not equal.

Some oversold readings happen during normal pullbacks.

Others happen during unstable volatility regimes where the market is moving fast, gaps are larger, and selling pressure can persist longer than expected.

That distinction matters.

The original version captures the broad edge, but it does not ask whether the environment is suitable for mean reversion.

That is what our improved version fixes.

It keeps the same core Larry Williams idea, but adds one extra layer of protection before allowing new trades.

I will show the exact rule and code later in the article.

For now, the important point is this:

The goal was not to increase trade count.

The goal was to improve trade quality.

Why the Improved Version Is Better

The improved version made less money on ES as a standalone strategy, but the quality of the return stream improved.

That is the key.

The original version produced more raw profit, but it also carried more pain.

The improved version reduced exposure and made the system cleaner.

Here is the before and after:

Original ES version

Net Profit: $302,082.50
Trades: 529
Win Rate: 73.91%
Profit Factor: 2.05
Average Trade: $571.04
Largest Loss: -$19,450
Intraday Drawdown: $46,712.50
Max Trade Drawdown: $28,287.50
Time in Market: 28.89%

Improved ES version

Net Profit: $171,832.50
Trades: 340
Win Rate: 72.94%
Profit Factor: 2.37
Average Trade: $505.39
Largest Loss: -$9,087.50
Intraday Drawdown: $19,950
Max Trade Drawdown: $12,800
Time in Market: 18.19%

At first glance, someone might say:

“The original made more money. Why is this better?”

Because portfolio building is not only about standalone net profit.

It is about quality of return.

The improved version:

Reduced trades from 529 to 340
Improved profit factor from 2.05 to 2.37
Reduced largest loss from -$19,450 to -$9,087.50
Reduced intraday drawdown from $46,712.50 to $19,950
Reduced max trade drawdown from $28,287.50 to $12,800
Reduced time in market from 28.89% to 18.19%

That is a cleaner strategy.

It gave up some raw profit, but it became a better portfolio component.

That is exactly the trade off I am looking for.

The Bigger Test

Once the improved version was built, I tested whether the Williams %R edge could contribute to a wider mean reversion portfolio. (Strategies that where included in the portfolio are listed below)

That is where it became more interesting.

The aim was not to create a portfolio full of identical strategies.

The aim was to combine strategies that all sit inside the mean reversion family, but trigger from different information.

Williams %R measures short-term closing location inside a recent range.

Other models in the portfolio look at different types of exhaustion, different market rhythms, and different pullback structures.

That is what matters.

If every strategy is basically the same, the portfolio is just one idea with more contracts.

If the strategies are different enough, the portfolio can become smoother than the individual parts.

The Williams %R model had low correlation to the other strategies in the portfolio.

That made it useful.

Not because it was perfect.

Because it fit.

The Final Mixed Strategy Portfolio

The final mixed strategy portfolio produced:

This is where the strategy becomes most useful.

On its own, Williams %R is good.

Inside the portfolio, it becomes more valuable.

The mixed portfolio produced $842,172.25 with a maximum drawdown of $22,364.

That gives a 37.66 Return/DD ratio.

For every $1 of maximum drawdown, the portfolio generated $37.66 of net profit.

That is the key.

The Williams %R edge did not need to be perfect.

It just needed to be profitable, robust, and different enough from the other systems to improve the portfolio.

That is what it did.

Why This Matters

The original Larry Williams idea was already good.

But the improved version is more useful for portfolio construction.

It does not just ask:

Is the market oversold?

It also asks:

Is this the type of environment where I actually want to buy mean reversion?

That second question made the strategy cleaner.

That is the real improvement.

Not more trades.

Not more complexity.

Better selectivity.

How the Improved Strategy Works

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

💬 Message to readers

This week’s issue is a Gold day trading model, and what makes it interesting is what it doesn’t do.

There is no trend filter. No regime filter. No higher timeframe confirmation. No stacked conditions trying to look clever.

It is a pure volatility capture strategy.

That is what I like about it.

Most systems try to filter the market to death before they allow a trade. This one takes the opposite route. It accepts that Gold has a habit of doing one thing very well: sitting quiet, then expanding hard. Instead of trying to predict when that move should happen, the strategy simply positions itself to catch it when it does.

And Gold is a very good market for that kind of idea. It is active enough, emotional enough, and explosive enough that you do not always need a complex filter to extract something useful. Sometimes the move itself is the edge.

The structure is also clean. One long entry concept. One hard stop. One trailing stop. Flat by the close.

That is exactly the kind of system I like in a portfolio. Not because it is flashy, but because it is simple enough to trust and robust enough to matter.

The other interesting piece here is the session template. This strategy runs on a custom 09:00 to 15:30 session. That means it can suit traders using day-rate margin, and it can also fit prop firms that only allow day session trading. Of course that comes with its own risk. Day-rate margin can disappear when volatility spikes, so you still need to respect that. But from a practical trading point of view, this is a very usable structure.

And the numbers are strong enough to take seriously. Since 2002, the strategy produced 877 trades, $129,230 net profit, a 1.80 profit factor, a 43.90% win rate, and 1,570.23% net profit to max closed-trade drawdown. That is a very respectable profile for a simple long-only intraday Gold system.

📖 The Idea

The idea is very simple.

Gold often spends time compressing, then suddenly expands. This strategy is designed to catch that expansion when it breaks out.

There is no entry filter deciding whether conditions are “good enough.” The strategy simply places a buy stop above price using recent volatility and lets Gold do the rest. If the market expands, it gets pulled into the trade. If not, nothing happens.

Once in the trade, risk is managed with a tighter ATR stop and then a much slower trailing ATR stop. Everything is closed by the end of the day.

That is the whole model. No prediction. No regime guessing. Just volatility expansion, risk control, and daily flattening. The code reflects exactly that structure: a long-only breakout entry above the current bar high plus an ATR-based offset, then a hard stop, a trailing stop and exits at 15:00.

🧠 Why this works for Gold

Gold is one of the better markets for a clean volatility breakout.

It can spend hours doing very little, then suddenly trend with real force. When that happens, the move is often clean enough that you do not need layers of filtering to capture it. You just need a sensible way to define the breakout and a sensible way to manage the risk afterward.

That is what this model is doing.

It is not trying to forecast direction in advance. It is waiting for Gold to expand and then joining the move once price proves it can break beyond its recent range.

🎯 Why I like it in a portfolio

I like this one because it does one job clearly.

It is not a mean reversion model. It is not a trend filter model. It is not a portfolio of conditions pretending to be a strategy. It is a straight volatility capture model for Gold.

That gives it a nice role inside a broader portfolio.

If you already run index systems, bond systems, or counter-trend ideas, this gives you something different. It is long only, session-based, and focused on expansion. That helps from a diversification point of view.

It is also practical. Because it runs on the 09:00 to 15:30 template, it is far more usable for traders who need to stay inside day-session rules. That opens the door for day-rate margin users and some prop-style environments. Again, that is not risk-free brokers can remove reduced margin in fast markets but from a workflow point of view it is a very tradable setup.

📈 Results of strategy since 2002

Since 2002, the strategy produced:

Net profit: $129,230.00
Profit factor: 1.80
Trades: 877
Win rate: 43.90%
Average trade: $147.35
Max closed-trade drawdown: $8,230.00
Return to max closed-trade drawdown: 1,570.23%
Trading period: 22 years, 7 months
Time in market: 1.09%

That is a strong profile for a simple, long-only Gold day model.

Results of strategy

Equity of strategy

The equity curve is what you want to see from this type of model. It is not relying on a handful of trades. It keeps stepping higher over a long sample.

🧪 Robustness testing

I trust a strategy only after it survives the usual stress tests.

Monte Carlo OHLC

Here we adjust the structure of the bars and re-run the strategy across 10000 simulations.

This is one of the more useful tests for a breakout model. If the edge disappears as soon as you disturb the bar structure, it is fragile. If it keeps holding together, that is a much better sign.

System Parameter Permutation test

We run many nearby combinations around the chosen settings.

This one is especially interesting here. The permutation work suggests the original result is somewhat overstated and that the median expectation is the more realistic number to focus on. That is exactly how it should be read. The good news is that the median result is still very solid, which is what matters most.

Alternative market test: CL

We apply the same logic to crude oil without redesigning it.

If the results are still acceptable, that tells us the idea is broader than just one market and less likely to be overfit to GC alone.

Alternative market test: HO

We apply the same logic to heating oil without redesigning it.

Again, we are not looking for perfection here. We are looking for evidence that the concept has legs outside one symbol.

Alternative market test: SI

We apply the same logic to silver without redesigning it.

That is a particularly relevant test because silver is related to Gold but still behaves differently enough to tell us something useful.

Alternative market test: NQ

We apply the same logic to NQ without redesigning it.

If it still produces something acceptable, that is another sign the core idea is broader than one market.

Alternative market test: BTC

We apply the same logic to BTC without redesigning it.

This is not because BTC and Gold are the same. It is because both can produce strong volatility expansions, and it is useful to see whether the breakout logic can travel.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

Most traders look for the next strategy.

I think the better question is this:

What happens when you start combining proven edges that behave differently from each other?

That is where this test became interesting.

I started with a simple mean reversion idea called IBS, or Internal Bar Strength. On its own, it is one of the cleanest ways to measure whether a market closed weak or strong relative to its own daily range.

Then I tested it across the major index futures.

Then I added it to a wider portfolio of strategies from previous research:

The goal was not to find the highest net profit strategy in isolation.

The goal was to create the best risk-adjusted portfolio.

That distinction matters.

A strategy can look amazing on its own and still be a poor portfolio component if it overlaps too heavily with what you already trade.

A strategy can also look simple on its own and become extremely valuable when added to the right mix of systems.

That is what happened here.

What IBS is measuring

IBS stands for Internal Bar Strength.

It measures where the market closes inside its own daily range.

If the market closes near the low of the day, IBS is low.

If the market closes near the high of the day, IBS is high.

That is the entire idea.

It is not trying to predict earnings.

It is not using macro data.

It is not forecasting next month.

It is asking one simple question:

Did the market close near the bottom of today’s range or near the top?

That matters because equity index futures have a long history of short-term mean reversion. When the market closes extremely weak, especially in a broad index, the next move often has a tendency to snap back.

Not always.

Nothing works always.

But as a behavioural tendency, it makes sense.

When price closes near the low of the day, it often reflects late-session fear, liquidation, forced selling, or short-term exhaustion. The next session can then see relief, short covering, or dip buying.

That is why IBS is such a useful concept.

It reduces the day to something very simple:

Did the market close near the bottom of today’s range or near the top?

Why I tested IBS across ES, NQ, and YM

One of the first things I want to know with any index strategy is whether the edge is isolated or broad.

If it only works on one contract, I get cautious.

If it works across ES, NQ, and YM, I get more interested.

The reason is simple.

ES, NQ, and YM are different instruments, but they are all major US equity index futures. If a short-term behavioural edge appears across all three, it suggests the model is capturing something broader than symbol-specific noise.

That is what happened here.

IBS worked across all three.

The ES test on a single contract no compounding produced:

Total Net Profit: $292,695
Total Trades: 765
Win Rate: 67.32%
Profit Factor: 1.72
Average Trade: $382.61
Largest Win: $11,400
Largest Loss: -$15,812.50
Trade Close Drawdown: $22,650
Intraday Drawdown: $43,487.50
Percent Time in Market: 36.38%
Annual Rate of Return: 4.91%

That is a strong standalone result.

But the important part is not just the total profit.

The important part is the structure.

The strategy has a high win rate, a decent average trade, and enough trades to give the sample some weight. It is not a 40-trade backtest pretending to be robust.

Then I tested NQ.

NQ was the strongest version in raw dollars.

That makes sense. Nasdaq futures generally carry more volatility, which means the same type of mean reversion effect can produce larger dollar swings.

Then I tested YM.

Again, profitable.

Again, similar behaviour.

That matters.

IBS is not just an ES curve fit. The same broad idea shows up across three major equity index futures.

The IBS only portfolio

After testing the instruments individually, the next step was to combine them.

The IBS-only portfolio across ES, NQ, and YM produced:

This is a solid result.

A 16.22 Return/DD ratio is strong.

Return/DD simply means how much profit the system produced for every dollar of maximum drawdown.

So a 16.22 reading means the IBS portfolio generated $16.22 of net profit for every $1 of maximum drawdown.

That is a very good number.

But there was a problem.

The IBS-only portfolio spent a lot of time in the market.

It was profitable, but it was still one family of ideas. It was the same concept applied across multiple index futures.

That is useful.

But it is not the final form of the portfolio.

The real question was:

Can IBS improve a broader mean reversion portfolio when combined with other edges?

Building the optimised portfolio

The next step was to add IBS into a group of strategies from previous articles and select the best combination based on risk adjusted performance.

The final portfolio used four different strategy families:

RSI2
A short-term oversold mean reversion model.

Turnaround Tuesday
A calendar-based behavioural reversal model.

IBS
A daily close-location mean reversion model.

Connors %B
A volatility-stretch pullback model.

These are all mean reversion strategies, but they do not trigger from the same information.

That is the whole point.

RSI2 looks at short-term momentum exhaustion.

Turnaround Tuesday looks at the weekly behavioural rhythm.

IBS looks at where the market closes inside the daily bar.

%B looks at where price sits inside a recent volatility envelope.

They are related, but they are not identical.

That is how you build a portfolio that is more than just one idea repeated four times.

Why correlation matters

Before combining strategies, I want to see how much they overlap.

If every strategy wins and loses at the same time, the portfolio is not diversified. It is just leveraged exposure to one idea.

The correlations were low across most of the portfolio components.

That is exactly what I want to see.

The strategies are not perfectly independent, but they are not moving as one block either.

The IBS strategy in particular had very low correlation to the other components.

That is what made it valuable.

Not just because it made money on its own.

Because it added a different return stream to the portfolio.

The final optimised portfolio

The optimised portfolio produced the following:

This is the result that stood out.

The portfolio made $710,946 with a maximum drawdown of only $17,183.

That gives a 41.37 Return/DD ratio.

That is not normal.

For every $1 of max drawdown, the portfolio generated $41.37 in net profit.

That is the kind of number you only get when the portfolio components are working together properly.

Not just making money.

Not just looking good separately.

Working together.

Why this portfolio is stronger than the individual parts

This is the main lesson.

The IBS portfolio by itself was good.

The Turnaround Tuesday portfolio by itself was good.

The Connors %B model by itself was useful.

The RSI2 model added another short-term mean reversion component.

But the best result came from combining them.

That is how systematic portfolios improve.

Not by trying to find one perfect strategy.

By stacking multiple imperfect edges that complement each other.

The individual systems each have weaknesses.

IBS can spend a lot of time in the market.

Turnaround Tuesday is selective and calendar-based.

%B is more specific to volatility stretch.

RSI2 can go through periods where oversold becomes more oversold.

But together, the combined portfolio becomes much stronger.

The weaknesses do not disappear.

They get diluted.

The strengths compound.

That is the real edge.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

Most portfolio upgrades do not come from discovering some completely new market secret.

They come from adding a strategy that behaves differently enough from what you already trade, while still fitting the same philosophy.

That was the appeal of adding the Connors %B mean reversion strategy to the existing Turnaround Tuesday portfolio.

For those who missed it, that original two-strategy portfolio is covered here:

That portfolio was already strong. It combined a filtered Turnaround Tuesday model on ES with a second daily mean reversion model on NQ. The result was $272,827.25 in net profit, a 2.26 profit factor, and a 16.82 Return/DD ratio.

Then I added one more strategy.

Not another Tuesday model.

Not another close to close calendar pattern.

A Larry Connors pullback strategy from 2009 that turned out to be a very good fit for the portfolio once I stripped it back to the long side only.

That single addition changed the portfolio materially.

And there is another reason this strategy deserves attention.

It was published in 2009.

That means from 2009 onward, the strategy has effectively been living in the real world as a public out of sample test. The market has had every chance to crowd it, trade it, weaken it, or arbitrage it away. Yet the strategy still held up well enough to improve the portfolio all the way through 2026.

That matters.

A lot of backtests look good before publication. Far fewer continue to hold up after the idea is already out in the open.

Before getting into what it did to the combined equity curve, it is worth understanding what kind of behaviour this strategy is actually trying to exploit, because the logic is very different from Turnaround Tuesday.

What this strategy is actually doing

This strategy comes from Larry Connors and Cesar Alvarez and was published in High Probability ETF Trading in 2009.

At its core, it is trying to exploit a simple market tendency:

Strong markets do not move up in a straight line.

Even during healthy uptrends, price gets stretched, traders panic, short-term selling becomes overdone, and the market snaps back.

That is the behaviour this model is trying to isolate.

It is not trying to buy strength.

It is not trying to predict breakouts.

It is not trying to hold for months.

It is looking for an uptrend that gets temporarily stretched lower, then betting on the rebound.

That makes it a very different animal from Turnaround Tuesday.

Turnaround Tuesday is a short-duration behavioural edge built around the weekly emotional rhythm of the market.

This Connors model is a trend-filtered pullback strategy. It cares about whether the market is in a healthy broader uptrend and whether the recent downside move has become sufficiently stretched.

That difference is exactly why it is useful in a portfolio.

It is still mean reversion.

But it is a different kind of mean reversion.

That means different entries, different holding patterns, different clusters of trades, and ideally a smoother combined result when paired with other systems.

Why the 2009 publication date matters so much

This is one of the most valuable parts of the whole test.

The strategy was published in 2009. That means the period from 2009 to 2026 is not just more backtest history. It is a long stretch of public, real-world, post-publication data.

That matters because once an idea is in a book, it is no longer hidden.

From that point on, anyone can trade it.

Funds can test it.
Retail traders can copy it.
Researchers can publish variations of it.
And the market has plenty of time to weaken the edge if it is fragile.

So when a strategy still holds up after publication, that tells us something important.

It does not guarantee the future.

But it does suggest the edge was not just a lucky historical artifact that only worked before anyone knew about it.

That is a big reason I take this result seriously.

A strategy that looked good only up to 2009 would be far less interesting.

A strategy that was published in 2009 and still improves a portfolio through 2026 is much more valuable.

It has already survived a long public out of sample window through very different market conditions:

post-GFC recovery, QE, low-volatility bull markets, the 2020 crash, the 2022 bear market, and everything since.

That kind of durability is rare, and it is worth respecting.

The standalone long-only results

Tested on @ES.D daily bars, the standalone long-only strategy produced the following:

Total Net Profit: $61,875.00
Total Trades: 182
Win Rate: 77.47%
Profit Factor: 1.79
Average Trade: $339.97
Gross Profit: $139,850.00
Gross Loss: -$77,975.00
Largest Win: $6,600.00
Largest Loss: -$8,362.50
Max Consecutive Wins: 16
Max Consecutive Losses: 2
Trade Close Drawdown: $18,312.50
Max Intraday Drawdown: $29,137.50
Percent Time in Market: 6.55%
Annual Rate of Return: 1.73%

There are a few things I like immediately.

First, the 77.47% win rate is high. That does not automatically make a strategy good, but for a pullback system it tells you the exits are harvesting rebounds consistently.

Second, it is long only, which in this case turned out to be a strength. Equity index mean reversion is usually cleaner on the long side. Once I removed the short side, the strategy became a much better fit for the rest of the portfolio.

Third, it is not just another copy of Turnaround Tuesday.

That is where the correlation matters.

The pairwise correlations between the strategies are low:

ES.D %B Long Only vs V1 ES: -0.07
ES.D %B Long Only vs V3 NQ: -0.08
V1 ES vs V3 NQ: 0.27

That is exactly what I want to see.

The new strategy is still mean reversion, but it is being triggered by a different kind of market behaviour. That gives the portfolio another source of returns without simply stacking the same edge three times.

The portfolio before adding the new strategy

Before adding the long-only Connors model, the existing two-strategy portfolio looked like this:

That was already a serious portfolio.

A 16.82 Return/DD ratio is excellent.

For every $1 of maximum drawdown, the portfolio generated $16.82 in profit.

The Monte Carlo on the old version was also strong.

At the 95% confidence level, the old portfolio still showed:

Net Profit: $131,103.44
Max DD: $42,011.26
Return/DD: 5.38

So the old portfolio was not weak.

That is important.

The goal here was not to rescue something broken.

The goal was to improve something already robust.

What changed after adding the long-only strategy

After adding the long-only Connors model to the Turnaround Tuesday stack, the combined three-strategy portfolio produced:

Read those numbers carefully.

The new portfolio made $61,875 more net profit than the old one.

That is a gain of roughly 22.7% in total net profit.

And the most important part is this:

Max drawdown did not increase.

The old portfolio had a maximum drawdown of $16,218.

The new portfolio still had a maximum drawdown of $16,218.

So the added strategy lifted profit materially without increasing the worst historical drawdown at all.

That is exactly the kind of portfolio improvement you want.

Before and after

Here is the direct comparison.

The old portfolio produced:

Total Net Profit: $272,827.25
Total Trades: 439
Win Rate: 56.72%
Profit Factor: 2.26
Return / DD Ratio: 16.82
Max Drawdown: $16,218
Average Trade: $621.47

After adding the long-only Connors strategy, the new portfolio produced:

Total Net Profit: $334,702.25
Total Trades: 621
Win Rate: 62.8%
Profit Factor: 2.14
Return / DD Ratio: 20.64
Max Drawdown: $16,218
Average Trade: $538.97

So the actual change was:

Net Profit: up $61,875.00
Trades: up 182
Win Rate: up 6.08 percentage points
Profit Factor: down slightly from 2.26 to 2.14
Return / DD: up from 16.82 to 20.64
Max Drawdown: unchanged at $16,218
Average Trade: down from $621.47 to $538.97

There is an important lesson in those numbers.

The new portfolio is not better because every single metric improved.

They did not.

The average trade came down.

The profit factor dipped slightly.

But those trade-offs were more than offset by what matters more at the portfolio level:

1. Total profit increased meaningfully
2. Maximum drawdown did not increase
3. Return / DD improved from 16.82 to 20.64

That is the real upgrade.

This is why you do not judge a new strategy purely on how pretty it looks in isolation.

You judge it on what it does to the portfolio.

A strategy can have a lower average trade than your best existing model and still be a great addition if it adds profit, improves sequencing, and does not worsen the drawdown profile.

That is exactly what happened here.

Why the long-only version improved the portfolio

There are three reasons this makes sense.

1. It attacks a different kind of mean reversion

Turnaround Tuesday is short duration and calendar driven.

Connors %B is stretch driven.

One is built around the weekly emotional rhythm of the market.

The other is built around repeated short-term downside compression inside a broader uptrend.

Those are related, but they are not the same.

That gives the portfolio another source of entries that is not dependent on the same exact behavioural trigger.

2. The long side is where equity index mean reversion is strongest

This is the big one.

On stock index futures, the long side of mean reversion is usually cleaner than the short side.

Fear tends to overshoot faster than optimism.

Selloffs tend to become more emotional, more compressed, and more reflexive.

That makes upside snap-backs cleaner than trying to short stretched rallies.

Once the short side was removed, the strategy became a better fit for the rest of the portfolio.

Less noise.

Cleaner behaviour.

Better integration.

3. It improved the return stream without worsening the worst pain

The strongest evidence is in the drawdown and Monte Carlo numbers.

If adding a strategy boosts profit but also makes the total portfolio much more fragile, that is not a real improvement.

That did not happen here.

The portfolio got larger while holding the same max drawdown.

That is exactly what I want to see.

The new portfolio equity

The equity curve makes the improvement easier to see than the summary stats alone.

You can see the original Turnaround Tuesday components continuing to do their job, while the new long-only pullback model adds another stream of returns on top.

That is the kind of addition I want in a portfolio.

Not something that replaces the old edges.

Something that layers on top of them.

Monte Carlo after adding the long only strategy

This is where the update becomes hard to ignore.

The new portfolio Monte Carlo results came out as follows:

Now compare the 95% line to the old portfolio.

Old portfolio at 95%
Net Profit: $131,103.44
Max DD: $42,011.26
Return / DD: 5.38

New portfolio at 95%
Net Profit: $167,767.94
Max DD: $50,837.07
Return / DD: 5.54

That is a useful improvement.

Even after reshuffling the trade order into worse sequences, the updated portfolio still holds up better on both profit and Return / DD.

That tells me the added strategy is not just making the backtest look nicer.

It is improving the structure of the portfolio in a way that survives stress.

How the strategy works

What makes this model attractive is that it is simple, specific, and tied to a real market behaviour.

It does not try to buy every dip.

It does not try to catch falling knives in weak markets.

It only wants pullbacks that happen inside a broader uptrend, and it wants those pullbacks to become sufficiently stretched before taking action.

That combination is what keeps the model from being too broad.

It is not buying weakness in general.

It is buying temporary overextension inside strength.

That is an important difference, and it is one of the reasons it fits well alongside the Turnaround Tuesday models rather than simply duplicating them.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

💬 Message to readers

This week’s issue is a 15-minute ES counter trend strategy.

Most traders are taught the same basic idea. Buy when price is above the 200-day moving average. Sell when price is below it. Follow strength. Go with the trend.

This model does the opposite.

When the daily trend is down, it looks for a long. When the daily trend is up, it looks for a short. At first that sounds backwards. But for stock indices, that is often where the opportunity sits. The best long entries often come when the market is under pressure and oversold. The best short entries often come when the market is stretched and overbought.

That is the whole angle here. We are not trying to chase strength. We are trying to exploit the snapback that often follows it.

And importantly, this is not just an idea. It has been tested over a long sample and the numbers are strong. In the main test the strategy produced 1,482 trades, $197,087.50 net profit, a 1.64 profit factor, a 48.99% win rate, and a 2,335.85% return to max closed-trade drawdown over 28 years and 6 months.

That is the kind of profile that gets my attention.

📖 The Idea

The idea is simple.

Use the daily chart to decide whether the market is stretched in one direction. If the daily trend is down, look for a long on the intraday chart. If the daily trend is up, look for a short.

Then wait for a small shift in intraday behaviour before placing an order. Do not enter at market. Use a stop order away from price so the market has to prove it can turn.

Once in the trade, keep risk tight with a hard ATR stop and then hand the job to a wider trailing stop. Everything is closed by 15:00.

This is a counter trend model, but it is not reckless. It is trying to buy weakness only when there is evidence the weakness is exhausting, and sell strength only when there is evidence the move is running out.

🧠 Why this works for ES

This works because equity indices do not behave like commodities.

Indices have a natural upward drift over time, but just as important, they also have a habit of snapping back after sharp moves. Panic selling can create strong long opportunities. Euphoric buying can create strong short opportunities. That is especially true intraday, where emotion and positioning can push price too far before the market stabilises.

That is why I find this model so interesting. It does not chase strength once everything already looks comfortable. It looks for moments where price is likely stretched, then waits for enough confirmation before getting involved.

The broader context matters, and so does the intraday behaviour. This strategy combines both. It first decides when the market is in the kind of condition where a counter trend trade makes sense. Then it waits for price to show the move may be losing control before acting. Finally, it still makes the market prove itself before entry and manages risk tightly once in the trade.

That is really the edge here. It is not blindly fading weakness or strength. It is looking for exhaustion, waiting for a shift, and then controlling the risk.

🎯 Why I like it in a portfolio

I like this model because it gives you a very different behaviour profile from a standard breakout or trend strategy.

Most trend models buy strength and sell weakness. This one fades those moves inside a broader context. That means it can bring a different return stream into a portfolio rather than just cloning the same risk as your existing systems.

It is also very simple.

One core idea. One hard stop. One trailing stop. One end-of-day exit.

That matters. Simple systems are easier to trust, easier to test, and easier to keep trading through the rough patches.

📈 Results of strategy since 1987

Net profit: $197,087.50
Profit factor: 1.64
Trades: 1,482
Win rate: 48.99%
Average trade: $132.99
Max closed-trade drawdown: $8,437.50
Return to max closed-trade drawdown: 2,335.85%
Trading period: 28 years, 6 months
Time in market: 2.10%

That is a very strong result for a simple intraday counter trend model.

🧪 Robustness testing

I trust a strategy only after it survives the usual stress tests.

Monte Carlo Price

Here we adjust the price close data by a factor to see if the strategy was only going to work on that exact sequence of price close data.

If the strategy falls apart from small distortions in the price close, that is a bad sign. If it keeps holding together, that is a strong sign the edge is real.

Monte Carlo OHLC

Here we disturb the open, high, low, and close of the bars.

This is a harder test. If the strategy still behaves properly here, it is much less likely to be relying on one very specific bar structure.

[

Monte Carlo randomise trades

Here we randomise the trade sequence.

A strategy can look comfortable simply because the wins and losses came in a friendly order. This test asks what happens if that order changes. If the distribution still looks healthy, that gives more confidence in the edge.

Walk Forward Matrix

We optimise on one window and test on the next unseen window, repeatedly. If the strategy stays stable across rolling periods, that is a strong sign it adapts better

System Parameter Permutation test

We run many nearby combinations around the chosen settings.

If most of that neighbourhood still works, the chosen parameters are probably not just lucky.

In the result below, after 1,000 simulations, the Ret/DD comes in very close to the selected settings. That is a good sign we have not simply picked a lucky parameter set.

Alternative market test: NQ

We apply the same logic to NQ without redesigning it.

If the results still hold up, that tells us the idea is broader than just ES and less likely to be overfit to one market.

Alternative market test: YM

We apply the same logic to YM without redesigning it.

If the results still hold up, that tells us the idea is broader than just ES and less likely to be overfit to one market.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

Larry Williams found one of the simplest edges in the market.

Buy Monday’s close after a weak Monday.

Sell Tuesday’s close.

That is it.

No prediction.
No complicated model.
No intraday monitoring.
No market commentary required.

The idea became known as Turnaround Tuesday.

The logic is behavioural.

Markets often absorb fear after the weekend. News builds up. Institutions rebalance. Retail traders panic. Monday can become the emotional flush.

Then Tuesday arrives.

Shorts cover. Dip buyers step in. The fear trade unwinds.

The original idea is simple:

Buy: Monday’s close, but only if Monday closes below Friday’s close
Exit: Tuesday’s close
Market: @ES.D daily
Hold time: one day
Trades: maximum one per week

Why the One-Day Hold Is the Entire Edge

Most mean reversion strategies eventually get hurt because they hold too long.

A 3-day hold can become a 5-day problem.

A 5-day hold can become a bear market trap.

A bear market trend can grind a “cheap” entry into dust.

That is what happened to a lot of strategies in 2022.

They worked beautifully during the 2010s bull market, then collapsed when the regime changed.

Turnaround Tuesday avoids that problem because it is barely in the market.

You are not trying to call the bottom.

You are not trying to predict next week.

You are not trying to survive a trend.

You are exploiting one behavioural window:

Monday fear.

Tuesday relief.

Exit.

That short hold time is the protection.

You cannot get trapped in a long trend if you are only holding for roughly 24 hours.

That is why the concept has survived 2002, 2008, 2020, and 2022 better than many longer-hold mean reversion systems.

But survival does not mean perfection.

What the Original Looks Like Today

I tested the original Turnaround Tuesday idea on:

Instrument: @ES.D
Market: E-mini S&P 500 continuous daily data
Period: November 1997 to April 2026
Position size: 1 contract
Frequency: maximum 1 trade per week

Here are the actual TradeStation results:

This is not a bad strategy.

That matters.

The original version is not broken. It makes money. It has a logical behavioural foundation. It spends very little time in the market. It has survived multiple market regimes.

But the equity curve still tells an important story.

The early period is rough. The first part of the curve spends years chopping sideways and dipping below zero. It eventually recovers, but the first stretch is not clean.

The strategy finishes at $109,037.75, but the path is not as smooth as the final number makes it look.

The Real Problem With the Original

The original does not buy every Monday.

That would be too crude.

The true original only buys when Monday closes below Friday’s close.

So the edge already understands the basic idea:

Monday needs to be weak.

But that still leaves a problem.

It treats every down Monday the same.

A Monday down 0.03% gets a trade.

A Monday down 0.15% gets a trade.

A Monday down 1.20% gets a trade.

Those are not the same setup.

One is noise.

One is mild weakness.

One may be genuine fear.

That is where the original leaves money on the table.

The original captures the broad Turnaround Tuesday tendency, but it also takes too many marginal trades where there is not enough selling pressure to create a meaningful reversal.

That is why the average trade is only $187.67.

That is why the profit factor is only 1.48.

Again, that is not terrible.

But it is not where the research ends.

What Our Version Fixes

Our version does not discover Turnaround Tuesday.

Larry Williams already did that.

Our version improves the edge by asking a better question:

Which Turnaround Tuesday setups are actually worth taking?

That distinction matters.

The original version trades every down Monday.

Our version focuses on higher-quality reversal conditions, then combines two complementary versions into a portfolio.

Here is the difference:

That is the key.

Our portfolio made about 2.5x more total profit.

It did it with 142 fewer trades.

The average trade improved from $187.67 to $621.47.

The profit factor improved from 1.48 to 2.26.

And return-to-drawdown improved from roughly 7.39x to 16.82x.

Return-to-drawdown means this:

For every $1 of maximum drawdown the strategy had to survive, how much total profit did it produce?

The original produced about $7.39 for every $1 of drawdown.

The portfolio produced $16.82 for every $1 of drawdown.

That is the real improvement.

Not just more profit.

Better profit per unit of pain.

The public version shows the original edge.

Our version shows how we improved the trade selection, validated it across markets, combined it into a two-instrument portfolio, and stress-tested it with Monte Carlo.

That is where the real research is.

Original Turnaround Tuesday Code

Run this on @ES.D daily bars.

Inputs:
    ATRLength ( 14  ),
    ATRMult   ( 2.0 );

Variables:
    MondayDrop ( 0 ),
    StopAmt    ( 0 );

If DayOfWeek(Date) = 1 Then Begin
    MondayDrop = (Close[1] - Close) / Close[1];
    If MarketPosition = 0 AND MondayDrop > 0 Then Begin
        StopAmt = ATRMult * AvgTrueRange(ATRLength) * BigPointValue;
        SetStopLoss(StopAmt);
        Buy("TT Buy") This Bar on Close;
    End;
End;

If MarketPosition = 1 Then Begin
    If DayOfWeek(Date) = 2 Then
        Sell("TT Sell Tue") This Bar on Close;
    If DayOfWeek(Date) = 3 Then
        Sell("TT Sell Wed") This Bar on Close;
End;

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

💬 Message to readers

This week’s issue is a KAMA and ATR model for RTY on 15-minute bars.

At the centre of this strategy is KAMA, short for Kaufman Adaptive Moving Average. It was created by Perry Kaufman to solve a simple problem: normal moving averages are too slow when markets move cleanly and too reactive when markets get noisy. KAMA adjusts itself. This KAMA is configured to be deliberately cautious in trends and more reactive in choppy conditions, which filters out a lot of noise before a signal fires.

That matters a lot for RTY.

Russell is not a smooth market. It is rougher than ES, more erratic than YM, and it can change character quickly. A fixed moving average can get chewed up in that kind of tape. KAMA does a much better job because it adapts to the market in front of it. With these settings, KAMA is cautious during strong trends and more sensitive during choppy periods, which means signals only appear when price has genuinely organised itself.

The other thing I like here is how simple the structure is. One trend read. One entry. One hard stop. One trailing stop. Flat by the close. That is exactly the kind of system I want in a portfolio. Not something flashy. Something robust.

And the sample size matters. This strategy produced 3,923 trades over 24 years and 4 months. That is more than enough to take the results seriously. Even after including 1 tick slippage and commission, it still delivered $157,155.50 net profit, a 1.27 profit factor, a 46.27% win rate, and $40.06 average trade. That is the kind of profile I care about. Long history. Large sample. Simple logic. Robust behaviour.

📖 The Idea

The idea is very simple.

Use KAMA as the trend signal. When price crosses above it, prepare for a long. When price crosses below it, prepare for a short.

Once the signal appears, the strategy enters on the next bar. There is no attempt to overcomplicate the entry. The edge here comes from getting aligned with short-term direction early, then managing risk properly once in the trade.

From there, ATR takes over for the risk management. The model uses a hard ATR stop to define risk and a wider ATR trailing stop to stay with the move if it expands. Everything is closed at 15:00 so there is no overnight exposure.

That is the whole model. KAMA for direction. ATR for risk. Flat by the close.

🧠 Why this works for RTY

RTY tends to move in bursts. It can drift for a while, then suddenly expand and travel hard. That is the environment where this kind of model makes sense.

KAMA helps because it adapts to the quality of the move. When RTY is trending properly, KAMA becomes useful as a directional anchor. With these settings it is conservative by nature, so it does not fire signals constantly and keeps trade count manageable.

That matters because this strategy is not trying to predict reversals or catch every small move. It is simply trying to get aligned with short-term direction when the move starts to organise itself.

Then ATR takes over on the risk side. RTY volatility changes a lot, so fixed stops are a poor fit. ATR keeps the stop in line with the market and gives the trade room when the instrument is moving.

🎯 Why I like it in a portfolio

I like this one because it does one job clearly.

It is an intraday trend model for RTY. It is not trying to predict reversals. It is not trying to scalp every little movement. It waits for direction, gets in, and then manages the trade with adaptive risk.

That makes it useful in a portfolio. If you already run mean reversion, bonds, metals, or slower index systems, this gives you something different without just repeating the same exposure. It is also simple enough that you can actually stick with it.

📈 Results of strategy since 2001

This strategy has been tested from 2001 to 2026 and produced 3,923 trades over that period.

After including 1 tick slippage and commission, the strategy delivered:

Net profit: $157,155.50
Profit factor: 1.27
Win rate: 46.27%
Average trade: $40.06
Time in market: 6.26%
Max closed-trade drawdown: $8,109.50
Return to max drawdown: 1,937.92%

Without costs, the strategy improves to:

Net profit: $182,655.00
Profit factor: 1.32
Win rate: 46.78%
Average trade: $46.56
Max closed-trade drawdown: $7,805.00
Return to max drawdown: 2,340.23%

Results of strategy

Equity of strategy

I also want to note that 2001 to 2007 was left completely untouched during development.

We ran the model once and did not use that window in the build process.

That is about as close to live as you can get in historical testing, and the equity through that stretch matters far more than a polished in-sample result.

🧪 Robustness testing

Monte Carlo Price

Here we adjust the price data by a factor to see if the strategy was only going to work on that exact sequence of price data.

If the strategy falls apart from small distortions in price, that is a bad sign. If it still holds together, the edge is probably real.

Monte Carlo OHLC

Here we go one step further and perturb the open, high, low, and close structure of the bars.

This is a harder test than just changing the close sequence. If the model still behaves reasonably well, it tells us it is not relying on a very specific bar pattern to survive.

Monte Carlo randomise trades

Here we randomise the trade sequence to see what happens if the wins and losses come in a different order.

This matters because a strategy can look comfortable only because the historical trade order was favourable. If the equity still looks acceptable after randomisation, that is a strong sign.

Walk Forward Matrix

We optimise on one window and test on the next unseen window, repeatedly.

If the results stay reasonably stable across rolling periods, that is a strong sign the strategy adapts better than a curve fit model.

System Parameter Permutation test

We run many nearby combinations around the chosen settings.

If most of that neighbourhood works, the chosen parameters are probably not just lucky.

Alternative market test: EMD.D

We apply the same logic to EMD.D without redesigning it.

If the results are still acceptable, that tells us the idea is broader than just one market and less likely to be overfit to RTY alone.

Alternative market test: NQ.D

We apply the same logic to NQ.D without redesigning it.

If the results are still acceptable, that tells us the idea is broader than just one market and less likely to be overfit to RTY alone.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

💬 Message to readers

Improving a strategy is hard.

Most traders try to improve performance by adding more filters, more conditions, and more parameters. On the surface it looks smart. In reality, most of the time it is just curve fitting. The computer is simply finding tiny quirks in old data and dressing them up as improvements.

That is why genuine strategy improvement is so difficult.

A much better approach is to leave the core logic alone and instead understand when a strategy performs well and when it performs poorly. Once you know that, you can push harder when conditions are favourable and back off when conditions are not.

That is what this article is about.

Instead of changing the engine, we are changing how hard we press the accelerator.

📖 The Idea

The idea is simple.

We created a library of dynamic position sizing filters that add market context to an already robust strategy. Instead of asking, “Can I make the entry logic smarter?” we ask, “Should this strategy be aggressive right now, normal right now, or should it stand aside?”

That is a very different question.

For this framework, we built filters using two market context tools we like a lot: VIX and ADD market breadth.

The VIX is the market’s fear gauge. It tells you when volatility expectations are calm and when they are stressed. Some strategies love calm conditions. Others do better when volatility expands. If you know which environment fits your strategy, you can size up when the odds are better and reduce exposure when they are not.

The second input is $ADD, which measures how many NYSE stocks are advancing versus declining. In plain English, it tells you whether the market is moving with broad participation or whether a move is being carried by only a few names. That matters because broad, healthy participation often produces very different trading conditions from narrow, fragile participation.

So the goal here is not prediction.

The goal is to let market context control position size.

Sometimes that means trading normal size.
Sometimes that means doubling size.
Sometimes that means skipping the trade completely.

That is dynamic position sizing in its most practical form.

⚙️ The Setup

For this test, we used our previously published ES.D day trading strategy. It is already a strong and robust performer, which makes it a good candidate for dynamic position sizing. We are not trying to rescue weak logic here. We are trying to improve the deployment of a strategy that already has a real edge.

The setup is straightforward.

For this strategy:

• Data3 = $VIX.X

• Data4 = $ADD

• both are added as daily bars

• the chart must be set to local time, not exchange time

In my setup, local time on my PC is CME, so everything stays aligned properly. If you run this yourself, make sure you do the same, because time alignment is one of the easiest ways to accidentally break a multi-data strategy.

To install it, add Data3 ($VIX.X, 1440m) and Data4 ($ADD, 1440m) or change the snippit to suit Data2 or Data3 as subcharts to your strategy chart. Paste the DPS block into your strategy code before your entry orders. Add SizeMode(0) and SizeScale(2) to your inputs. Add the required DPS variables to your vars block. Replace your order quantity with FinalLots, then add:

FinalLots = Lots * MaxList(DynSize, 1);

and

If DynSize = 0 then OpenOrdersAllowed = False;

at the bottom of the DPS block.

Finally, set MaxBarsBack = 300 in Strategy Properties.

For the improved run shown in this article, the winning filter was Filter 4, which is the fixed VIX regime filter. In the snippet, Filter 4 uses hardcoded VIX bands to classify the environment as calm, normal, or stressed. At SizeScale = 2, VIX below 15 is treated as calm, 15 to 24 is treated as normal, and above 24 is treated as stressed. Calm conditions can size up. Normal conditions stay at base size. Stressed conditions are skipped.

That is important.

This filter is not trying to predict direction.
It is simply trying to answer one question:

How aggressive should this strategy be in this volatility regime?

📈 Results

Baseline results

The baseline version of the strategy is already strong.

It produced $178,125 in total net profit over 1,130 trades, with a 1.60 profit factor, a 51.15% win rate, and an average trade net profit of $157.63. Maximum intraday drawdown came in at $11,550, while trade-close-to-trade-close drawdown was $10,725.

That is exactly the kind of baseline you want before testing deployment filters. The strategy is already proven. We are not curve fitting a bad system into looking acceptable. We are testing whether a good system can be deployed more efficiently.

Improved results using Filter 4

With the dynamic position sizing block enabled using Filter 4, total net profit increased to $224,875 over 1,011 trades. Profit factor came in at 1.58, the win rate was 50.54%, and average trade net profit jumped to $222.43. Maximum intraday drawdown increased to $15,500, and trade-close-to-trade-close drawdown increased to $11,862.50.

The first thing that stands out is simple.

The strategy made more money from fewer trades.

That is exactly what you want to see.

What actually improved

Compared with the baseline, total net profit increased by $46,750, which is a 26.25% improvement. Trade count fell by 119 trades, which is a 10.53% reduction. Average trade net profit increased by 41.11%. Return on initial capital improved from 178.13% to 224.88%, while annualised return improved from 3.59% to 4.13%. Trade-close-to-trade-close efficiency also improved, with Net Profit as % of Drawdown rising from 1,660.84% to 1,895.68%.

That is a very clean result.

The filter did not improve performance by firing more often.
It improved performance by being more selective and more aggressive only when conditions were more favourable.

In other words, it improved deployment quality.

Why Filter 4 is improving results

This strategy appears to perform better when volatility is controlled, not chaotic.

That is what makes Filter 4 such a good fit.

When VIX is low to moderate, intraday structure tends to be cleaner. Breakouts behave more normally. Stops are less likely to get hit by panic-driven noise. The market still moves, but it moves in a more tradeable way.

When VIX is elevated, conditions change. Ranges become more violent. Reversals become sharper. Good-looking setups are more likely to fail simply because the market is unstable.

Filter 4 addresses that directly.

It does three simple things:

• it sizes up when VIX is calm

• it keeps size normal in acceptable conditions

• it stands aside when VIX is stressed

That is why the results make sense.

The strategy took fewer trades, but each trade was worth much more on average. That strongly suggests the filter is removing lower quality high-stress trades and concentrating risk into regimes where the edge is more reliable. The filter logic itself is intentionally simple: fixed VIX bands, no complexity, no fancy modelling, just volatility context controlling exposure.

The trade-off

This was not a free lunch across every metric.

Profit factor slipped slightly from 1.60 to 1.58.
Win rate dipped from 51.15% to 50.54%.
Intraday drawdown widened from $11,550 to $15,500.
Required account size also rose from $10,725 to $11,862.50 because the strategy is sometimes trading larger size.

That matters, and it should be acknowledged.

But those trade-offs are acceptable here because the important things improved:

• total net profit

• average trade value

• return on capital

• annual return

• trade-close deployment efficiency

That is exactly what dynamic position sizing is supposed to do.

It is not meant to make every single metric better.
It is meant to help you deploy capital more intelligently.

And that is exactly what happened here.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

Crude oil is one of the hardest markets to trade well over time.

That is even more true when risk rises, volatility expands, and the market becomes more sensitive to sudden repricing. Recent conflict driven energy disruption and Strait of Hormuz risk have pushed crude back into a more unstable, shock prone regime, with sharp moves in Brent and WTI alongside severe shipping disruption and a higher geopolitical risk premium.

That context matters.

Not because this is a macro article.

Because unstable markets expose weak systems fast.

A strategy can look fine in a calm regime. Then conditions change, trade quality slips, and the edge starts leaking out through a long series of average trades.

That is what makes this result interesting.

Tested on CL 15-minute data from 2001 to 2026, this strategy produced:

Net Profit: $110,850
Profit Factor: 1.5
Trades: 824
Max Drawdown: $17,410
Return on Initial Capital: 110.85%
Time in Market: 2.53%

Those numbers are strong.

What makes them more interesting is the profile behind them.

This is not a strategy built on constant exposure. It is selective, stays out of the market most of the time, and still puts together a strong long-run result.

That is the kind of structure worth paying attention to in crude oil.

Why this matters

Most weak CL systems do not fail all at once.

They erode.

The backtest still looks acceptable. The logic still fires. But the market has changed, and the strategy starts paying for it one mediocre trade at a time.

That is the real problem in crude oil.

The challenge is not finding something that worked for one stretch. The challenge is building something selective enough to avoid bad participation and durable enough to keep functioning when the regime shifts.

This one appears to do that well enough to deserve a closer look.

The first thing that stood out

The first thing worth noticing is the exposure profile.

Only 2.52% time in market across almost 25 years.

That immediately separates it from a lot of intraday systems. It is not trying to grind out returns through constant activity. It is waiting for more specific conditions.

That matters in CL, because overtrading is one of the fastest ways to destroy a decent idea in this market.

The equity curve looks believable

The equity curve is not perfectly smooth, which is a good sign.

Real futures strategies stall, recover, flatten out, and then step higher again. They reflect changing conditions instead of pretending every regime was friendly.

That is what shows up here.

There are pauses. There are drawdowns. But over the full sample, the system keeps rebuilding.

That makes the result more credible.

What makes this strategy interesting

At a high level, the architecture looks right for crude oil.

There is broader context framing the trade, an intraday trigger handling execution, and risk management doing enough work to keep the system from becoming overexposed to the market’s worst behavior.

That combination matters.

Purely local signals can work for a while in CL, but they often break down when the regime shifts. A strategy with more structure usually travels better.

That is one of the strongest features here.

This is a selective system

This is not a “more trades, more edge” strategy.

It is closer to the opposite.

The strategy appears to benefit from doing less, but under better conditions.

That is often exactly what crude oil demands, especially when the market is more reactive to war headlines, supply fears, and volatility shocks. Current disruption has made oil far more sensitive to supply risk, shipping interruptions, and abrupt changes in the risk premium.

More participation in CL does not automatically improve a system. Often it just means more friction, more noise, and more ways for a fragile strategy to decay.

This one looks built around selectivity first.

The robustness work matters

A polished backtest is easy to produce.

The real question is what happens when the strategy gets stressed.

That is where this becomes more interesting.

The Monte Carlo distribution stayed constructive.
The walk-forward result passed.
The strategy parameter performance held together well enough to suggest the edge is not sitting in one tiny lucky setting.

That does not guarantee future performance.

It does suggest the system has structural integrity.

The parameter stability result is especially important.

Fragile research usually falls apart the moment the settings are disturbed. Stronger research still looks like itself even when the parameters move around.

That is a much better sign.

Why this is worth studying

Up to this point, the article has only covered the external profile:

the long test window,
the selective exposure,
the equity curve,
and the robustness work.

But the real value is in the internal structure.

That is where the strategy becomes useful, not just interesting. Once the logic is broken down clearly, the result starts to make sense. The equity curve stops being just a picture and starts looking like the output of a specific design.

That is where the deeper work begins.

Read more

Research and education only. Results are hypothetical and based on backtests and simulations. Past performance does not predict future results. Futures and derivatives involve significant risk. Test on your own data, costs, execution, and infrastructure before trading.

Most traders don’t ruin their systems at the entry.

They ruin them at the exit.

You can be right on direction…

and still butcher the trade.

That’s what dumb exits do.

They don’t know when trend strength is still alive.

They just pull the plug on schedule.

So I tested a different idea:

Use ADX for exits, not entries.

Not to ask:

Should I take this trade?

But to ask:

Does this trade still deserve to be held?

That one shift changed everything.

On a simple long-only NQ breakout, the baseline produced a 3.42 return-to-drawdown ratio.

Once ADX started managing the exit, that climbed to 5.63.

And when I pushed the framework further, it reached 6.12.

Same market.

Same core entry idea.

Smarter exit logic.

That’s what this article is about.

Most traders are solving the wrong problem

If you’ve built systems for any length of time, you’ve probably felt this.

You catch the move.

You survive the noise.

Then you either cut the winner too early…

or sit there giving back open profit because your exit has no feel for whether the trend is still healthy.

That’s the part that stings.

Not being totally wrong.

Being almost right and still walking away with less than you should.

Most traders respond by hunting for a better entry.

Another filter.

Another setup.

Another layer of confirmation.

But a lot of the time, the entry isn’t the real leak.

The leak is what happens after you get in.

That’s why I wanted to test ADX in a different role.

The contrarian angle

ADX is usually treated like a gatekeeper.

Trade only when trend strength is present.

That’s the standard play.

And sometimes it works.

But I think a lot of traders are using it too early in the decision chain.

Because once you’re already in the trade, the question changes.

You no longer need help deciding whether to participate.

You need help deciding whether the trade still deserves your capital.

That’s a different job.

Less prediction.

More trade management.

And for breakout systems, that distinction matters.

A lot.

What ADX is really useful for

ADX measures trend strength.

Not direction.

That’s what makes it so interesting here.

A good exit tool does not need to predict the next bar.

It does not need to forecast.

It does not need to tell you the future.

It just needs to answer one practical question:

Is there still enough strength behind this move to justify staying in?

That’s it.

That’s the entire idea.

Instead of using ADX as a front-door filter, I used it like a pressure gauge during the trade.

If the move still had strength, the trade could breathe.

If that strength deteriorated enough, the position no longer deserved to be held.

Simple concept.

Very different outcome.

The baseline

To test it properly, I started with a clean baseline.

Nothing fancy.

Long-only NQ breakout.

60-minute chart.

Buy above the highest high of the last 5 bars.

Exit after 14 bars if nothing else closes the trade first.

Commission included.

One tick of slippage included.

That baseline produced:

Net Profit: $287,817
Profit Factor: 1.13
Trades: 2,445
Largest Trade Drawdown: $23,940
RET/DD: 3.42

That’s a solid control.

Not spectacular.

Not terrible.

Exactly the kind of simple base model worth improving.

What changed

I didn’t change the market.

I didn’t change the core breakout concept.

I didn’t pile on ten new entry conditions.

I changed the exit logic.

That was the whole experiment.

Can ADX do a better job managing the life of the trade than a fixed hold alone?

The answer was yes.

And not by a little.

Once ADX started deciding whether trend strength still justified staying in the trade, the system improved materially.

The first major improvement took the strategy to:

Net Profit: $319,355**
Profit Factor: 1.16
Trades: 3,300
Largest Trade Drawdown: $20,045
RET/DD: 5.63

That gets my attention for a few reasons.

More net profit.

Better return relative to drawdown.

Lower largest trade drawdown.

Same entry.

Better trade management.

That’s exactly the kind of improvement I want to see in strategy research.

Not some Frankenstein rebuild.

Just a cleaner way to manage an open position.

The first layer of implementation

I also wanted this article to be practical, not just conceptual.

So below is the exact snippet used for the initial ADX exit logic. Just add this to your strategy and run your own experiments.

ADX exit snippet

//------------------------------------------------------------------
// ADX Exit Filter (Filters 1-4)
// Add inputs, vars, calc, and filter block to your strategy
//
// ExitFilterMult optimization range: 0.5 to 2.0 step 0.1
// Each filter's base constant is calibrated so that this range
// produces meaningful variation without degenerate edge cases.
//------------------------------------------------------------------

// ================================================================
// INPUTS — add to your inputs section
// ================================================================
// ExitFilterSelect(0),   { 0=off, 1=ADX rising, 2=ADX floor,
//                          3=ADX strong-trend floor, 4=DI+ above DI- }
// ExitFilterMult(1.0),   { 0.5=lenient  1.0=default  2.0=strict }
// ADXPeriod(14);

// ================================================================
// VARS — add to your vars section
// ================================================================
// TrueHigh(0),
// TrueLow(0),
// DM_Plus(0),
// DM_Minus(0),
// SmoothedDMPlus(0),
// SmoothedDMMinus(0),
// SmoothedTR(0),
// Prev_SmoothedDMPlus(0),
// Prev_SmoothedDMMinus(0),
// Prev_SmoothedTR(0),
// DI_Plus(0),
// DI_Minus(0),
// DX(0),
// ADX_Wilder(0),
// Prev_ADX(0),
// ExitFilterOK(0);

// ================================================================
// ADX CALCULATION — place before filter block
// ================================================================
TrueHigh  = MaxList(High, Close[1]);
TrueLow   = MinList(Low,  Close[1]);

if High - High[1] > Low[1] - Low then
    DM_Plus = MaxList(High - High[1], 0)
else
    DM_Plus = 0;

if Low[1] - Low > High - High[1] then
    DM_Minus = MaxList(Low[1] - Low, 0)
else
    DM_Minus = 0;

if CurrentBar <= ADXPeriod then begin
    SmoothedDMPlus  = Prev_SmoothedDMPlus  + DM_Plus;
    SmoothedDMMinus = Prev_SmoothedDMMinus + DM_Minus;
    SmoothedTR      = Prev_SmoothedTR      + (TrueHigh - TrueLow);
end else begin
    SmoothedDMPlus  = Prev_SmoothedDMPlus  - (Prev_SmoothedDMPlus  / ADXPeriod) + DM_Plus;
    SmoothedDMMinus = Prev_SmoothedDMMinus - (Prev_SmoothedDMMinus / ADXPeriod) + DM_Minus;
    SmoothedTR      = Prev_SmoothedTR      - (Prev_SmoothedTR      / ADXPeriod) + (TrueHigh - TrueLow);
end;

Prev_SmoothedDMPlus  = SmoothedDMPlus;
Prev_SmoothedDMMinus = SmoothedDMMinus;
Prev_SmoothedTR      = SmoothedTR;

if SmoothedTR <> 0 then begin
    DI_Plus  = 100 * SmoothedDMPlus  / SmoothedTR;
    DI_Minus = 100 * SmoothedDMMinus / SmoothedTR;
end;

if DI_Plus + DI_Minus <> 0 then
    DX = 100 * AbsValue(DI_Plus - DI_Minus) / (DI_Plus + DI_Minus);

if CurrentBar <= ADXPeriod then
    ADX_Wilder = (Prev_ADX * (CurrentBar - 1) + DX) / CurrentBar
else
    ADX_Wilder = (Prev_ADX * (ADXPeriod - 1) + DX) / ADXPeriod;

Prev_ADX = ADX_Wilder;

// ================================================================
// ADX EXIT FILTER BLOCK  (1-4)
// Place before entry logic
// ================================================================
ExitFilterOK = 1;

if ExitFilterSelect > 0 and MarketPosition > 0 then begin

    // 1: ADX rising — hold while trend momentum is strengthening bar-over-bar
    //    Exit when ADX ticks down or goes flat
    //    ExitFilterMult has no effect on this filter (pure direction check)
    if ExitFilterSelect = 1 then begin
        if ADX_Wilder <= ADX_Wilder[1] then ExitFilterOK = 0;
    end;

    // 2: ADX above floor — hold while trend is strong enough to trade
    //    Base = 15  |  Mult 0.5 = floor 7.5 (lenient)
    //               |  Mult 1.0 = floor 15  (default)
    //               |  Mult 2.0 = floor 30  (strict)
    if ExitFilterSelect = 2 then begin
        if ADX_Wilder < 15 * ExitFilterMult then ExitFilterOK = 0;
    end;

    // 3: ADX above strong-trend floor — stricter version of filter 2
    //    Base = 20  |  Mult 0.5 = floor 10 (lenient)
    //               |  Mult 1.0 = floor 20 (default)
    //               |  Mult 2.0 = floor 40 (strict)
    if ExitFilterSelect = 3 then begin
        if ADX_Wilder < 20 * ExitFilterMult then ExitFilterOK = 0;
    end;

    // 4: DI+ above DI- — hold while bulls have directional edge
    //    Exit when bearish pressure overtakes bullish momentum
    //    ExitFilterMult has no effect on this filter (pure crossover check)
    if ExitFilterSelect = 4 then begin
        if DI_Plus <= DI_Minus then ExitFilterOK = 0;
    end;

end;

// Apply filter — place immediately after block above
if ExitFilterOK = 0 then begin
    if MarketPosition > 0 then Sell("ExitADXFilter") this bar on close;
end;

Even at this stage, the takeaway is already clear:

you do not need to reinvent the entry to materially improve the system.

You need the trade to stop acting blind after it opens.

Why this matters more than most traders think

This is the opportunity most traders miss.

They spend months trying to perfect the setup.

Meanwhile their exit is still using a kitchen timer.

That works fine when the move behaves perfectly.

But markets don’t care about your timer.

Some trades need more room.

Some trades are already dying before the hold period ends.

Some are still healthy when the default exit forces them closed.

That’s why exit logic matters so much.

A fixed exit is blind.

It cannot tell the difference between a strong trend and a tired one.

ADX can.

Not perfectly.

Nothing does.

But it can at least give the strategy a way to respond to the quality of the move instead of just the passage of time.

And that changes the shape of the equity curve.

The real takeaway

The biggest lesson here wasn’t that ADX is magic.

It’s not.

The lesson was that a simple strategy can get a lot stronger when the exit stops acting blind.

That’s the kind of improvement I care about.

Not more complexity.

Not more indicators.

Not more knobs to optimize.

Just a more intelligent answer to one question:

Should this trade still be open?

And when you frame the problem that way, ADX starts making a lot more sense on the exit side than most traders realise.

Where the jump gets interesting

Up to this point, you’ve seen the idea.

You’ve seen the baseline.

You’ve seen the first lift in performance.

And you’ve seen the initial implementation.

But the part that made this worth publishing was what happened when I pushed the framework further.

That version produced:

Net Profit: $353,971
Profit Factor: 1.17
Trades: 2,529
Largest Trade Drawdown: $23,940
RET/DD: 6.12

That’s where this stops being a neat concept…

and starts becoming something you can actually build with.

Because now we’re not talking about a small cosmetic improvement.

We’re talking about a meaningful upgrade in risk-adjusted performance.

And if you’ve ever watched a good trade slowly die because your exit was too dumb to adapt, this is the moment that should hit home.

Because the difference between a decent strategy and a strong one is often not the entry.

It’s whether your exit knows when strength is still alive…

and when it’s quietly dying.

Read more

We’ve previously published a 30-minute Gold strategy using ADX.

That strategy was allowed to enter overnight and could also enter late in the session, which means you could be exposed to overnight margin requirements and overnight risk.

This week’s strategy is different.

This is a pure intraday GC strategy.
Entries are intraday.
Exits are intraday.
All positions are flattened by the daily cut-off.

So if you want Gold exposure without overnight holding, this is the cleaner version.

The logic is simple:

Wait for Gold to quiet down.
Make price prove itself with a breakout.
Control risk with volatility-based exits.
Get flat by the end of the session.

That is the kind of system I like in a portfolio. Simple rules. Clear job. No unnecessary complexity.

Gold fits well here because it tends to coil, then expand hard. This model is built to catch that expansion phase while avoiding overnight exposure.

📖 The Idea

This is a low-volatility breakout strategy for GC on 15-minute bars.

It looks for:

• low ADX

• price in the right location relative to a long term average price

• a breakout through a meaningful range

That combination matters.

Low ADX tells us the market is relatively quiet.
Average price gives directional context.
The breakout makes price prove the move is real.

Once in, the trade is managed with:

• an initial volatility stop

• a profit target

• a trailing stop

Only one trade per day is allowed.
Everything is closed by 15:00.
No overnight holds.
No weekend exposure.

🧠 Why this works for GC

Gold is not a market I want to trade with random intraday noise filters and endless indicators.

It tends to reward a cleaner approach.

A lot of its best moves come after periods of compression. That is why the setup starts with a low-volatility condition rather than a high-momentum one.

Then instead of predicting direction, the strategy waits for price to break out of a large lookback range.

So the workflow is:

1. market quiets down

2. directional context is checked

3. breakout order is placed

4. volatility handles the exits

That is a strong structure for GC because Gold can go from dead to fast very quickly.

And because this version is intraday only, it avoids a key issue from the earlier published model: you are not carrying positions into overnight margin and gap risk.

🎯 Why I like it in a portfolio

It does a different job.

Your portfolio should not be made up of ten versions of the same trend-following strategy. You want systems entering from different conditions and managing risk in different ways.

This one is useful because it is:

• intraday only

• compression to expansion

• breakout based

• volatility managed

• limited to one trade per day

That makes it a clean complement to slower swing systems or models that are willing to hold overnight.

📈 Results of strategy since 2002

Results of strategy

Equity of strategy (15-minute bars)

I also want to note that 2002-2007 was left completely untouched during development.

We ran the model once and did not use that window in the build process.

That is about as close to live as you can get in historical testing, and the equity through that stretch matters far more than a polished in-sample result.

🧪 Robustness testing

I trust a strategy only after it survives the usual stress tests.

Monte Carlo parameters

We vary the parameters around the chosen values to test stability.

If nearby settings still perform, the strategy is likely robust rather than dependent on one lucky combination.

Monte Carlo Price

Here we adjust the price data by a factor to see if the strategy was only going to work on that sequence of price data.

Walk Forward Matrix

We optimise on one window and test on the next unseen window, repeatedly.

If the results stay reasonably stable across rolling periods, that is a strong sign the strategy adapts better than a curve fit model.

System Parameter Permutation test

We run many nearby combinations around the chosen settings.

If most of that neighbourhood works, the chosen parameters are probably not just lucky.

Alternative market test: SI

We apply the same logic to Silver without redesigning it.

If the results are still acceptable, that tells us the idea is broader than just one market and less likely to be overfit to GC alone.

Alternative market test: HG

We apply the same logic to Copper without redesigning it.

If the results are still acceptable, that tells us the idea is broader than just one market and less likely to be overfit to GC alone.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

You do not need six indicators.
You do not need more optimisation.
And you definitely do not need another system that looks incredible until it meets real markets.

What you need is something much harder to find:

A simple edge that still behaves like an edge after everyone already knows about it.

That is why Larry Connors’ RSI(2) strategy still matters.

Not because it is flashy.
Not because it is new.
But because it solves a real problem:

How do you buy weakness in an uptrend without turning every dip into a disaster?

That is where this setup earns its reputation.

It is simple.
It is public.
And it gives you a strong baseline to build from.

Why this strategy still matters

Larry Connors did not invent RSI.

J. Welles Wilder created RSI.

What Connors did was turn it into one of the best-known short-term mean reversion frameworks traders still use today:

• buy short-term weakness

• only do it in the right broader trend

• exit on the snapback

That is why RSI(2) is still worth studying.

Not because it is complex.

Because it is robust.

It is built on a behaviour that still exists:

• markets trend

• pullbacks happen fast

• fear overshoots

• price snaps back toward the mean

The indicator is not the edge.

The behaviour is the edge.

Connors original RSI(2) strategy

The original logic is simple:

• long when price is above the 200-day moving average and RSI(2) is oversold

• short when price is below the 200-day moving average and RSI(2) is overbought

• exit longs when price closes above the 5-day moving average

• exit shorts when price closes below the 5-day moving average

That is it.

No fancy regime model.
No stack of filters.
No curve fit complexity.

And importantly, Connors’ original rules did not use a stop.

To stay true to the original research, that is how I tested the baseline.

That said, I have added an optional fixed-dollar stop feature to the code. So if you are not comfortable trading without a stop, you can enable it and define your own dollar risk per contract.

I also used 255 reserved bars in testing so all calculations had enough history to initialise properly.

{ ==============================================================
  LARRY CONNORS RSI2 STRATEGY - EasyLanguage for TradeStation

  TradeDirection Input:
    1 = Long Only
    2 = Short Only
    3 = Both Long and Short

  Original Rules:
    Trend Filter : Price vs 200-bar SMA
    Long Entry   : Close > 200 SMA AND RSI(2) < OversoldLevel
    Short Entry  : Close < 200 SMA AND RSI(2) > OverboughtLevel
    Long Exit    : Close > 5-bar SMA
    Short Exit   : Close < 5-bar SMA
    Stops        : Optional - Connors used none

  Stop is a fixed dollar amount per contract.
  e.g. StopAmt = 1000 means stop is $1000 from entry.
  Set UseStop = False to match Connors original rules.
============================================================== }


{ ── Inputs ───────────────────────────────────────────────── }
Input: RSI_Period      (2);
Input: OversoldLevel   (10);
Input: OverboughtLevel (90);
Input: TrendMA_Period  (200);
Input: ExitMA_Period   (5);
Input: TradeDirection  (3);      { 1=Long Only, 2=Short Only, 3=Both }
Input: UseStop         (False);  { Keep False to match Connors original }
Input: StopAmt         (1000);   { Fixed dollar stop per contract }

{ ── Vars ─────────────────────────────────────────────────── }
Var: MyRSI      (0);
Var: TrendMA    (0);
Var: ExitMA     (0);
Var: AllowLong  (0);
Var: AllowShort (0);

{ ── Calculations ────────────────────────────────────────── }
MyRSI   = RSI(Close, RSI_Period);
TrendMA = Average(Close, TrendMA_Period);
ExitMA  = Average(Close, ExitMA_Period);

{ ── Trade Direction ─────────────────────────────────────── }
AllowLong  = 0;
AllowShort = 0;
If TradeDirection = 1 Or TradeDirection = 3 Then AllowLong  = 1;
If TradeDirection = 2 Or TradeDirection = 3 Then AllowShort = 1;

{ ── Long Entry ──────────────────────────────────────────── }
If AllowLong = 1
And Close > TrendMA
And MyRSI < OversoldLevel
And MarketPosition <= 0 Then
    Buy("RSI2 Long") Next Bar at Market;

{ ── Long Exit ───────────────────────────────────────────── }
If MarketPosition = 1
And Close > ExitMA Then
    Sell("Exit Long") Next Bar at Market;

{ ── Short Entry ─────────────────────────────────────────── }
If AllowShort = 1
And Close < TrendMA
And MyRSI > OverboughtLevel
And MarketPosition >= 0 Then
    SellShort("RSI2 Short") Next Bar at Market;

{ ── Short Exit ──────────────────────────────────────────── }
If MarketPosition = -1
And Close < ExitMA Then
    BuyToCover("Exit Short") Next Bar at Market;

{ ── Fixed Dollar Stop (Optional) ────────────────────────── }
{ Connors used no stops - set UseStop = False to match original }
{ SetStopLoss monitors intrabar and takes dollar amount directly }
If UseStop Then
    SetStopLoss(StopAmt);

Original RSI(2) Strategy on ES.D

Market: ES.D
Reserved bars used: 255
Execution: Next bar at market
Position sizing: 1 Contract
Stops: Optional fixed-dollar stop included in code, but disabled by default to match Connors’ original rules

Equity curve since 1997

Largest Valley / Max Pain (Open Drawdown)

Results since 1997

What matters

This baseline gives us the benchmark.

If the original version still holds up after being public for years, then the underlying behaviour is still alive.

And that means we are not looking for a whole new strategy.

We are looking for a smart improvement.

The pro version: adding a VIX filter

This is where the strategy gets more interesting.

The raw RSI(2) setup treats every oversold signal the same.

That is the weakness.

Because not all pullbacks are equal.

Some are healthy dips that snap back fast.
Some are just noise.
And some happen in unstable conditions where mean reversion can keep failing.

That is why the VIX filter is so effective.

It helps us stop asking only:

“Is price stretched?”

And start asking:

“Is this the kind of volatility regime where this stretch is actually worth buying?”

Why this VIX filter works so well

For the pro version, the filter is simple:

Only take trades when the VIX condition is met.

This filter is not trying to buy maximum panic.

It is doing the opposite.

It is keeping the strategy out of the more unstable volatility regimes where oversold can stay oversold and mean reversion can break down.

That is why it works so well.

The original RSI(2) already finds stretched price action.

The VIX filter improves trade selection by saying:

• yes to pullbacks in a healthier, more stable volatility regime

• no to pullbacks in messy, high stress conditions where drawdowns tend to get worse

This is a very important distinction.

A lot of traders assume mean reversion always gets better when fear gets bigger.

That is not always true.

Sometimes extreme fear is not opportunity.

Sometimes it is instability.

This VIX filter helps avoid that.

Why this matters

It improves the strategy without changing its DNA.

It does not replace the edge.

It refines where the edge is allowed to operate.

RSI(2) Pro Version on ES.D

By adding the VIX filter to the original strategy, the improvements were meaningful:

• Profit factor increased from 2.15 to 2.99
  That is a 39.1% increase

• Average trade increased from $512 to $572
  That is an 11.7% increase

• Pain-to-gain / return to max drawdown increased from 946% to 1939%
  That is a 105.0% increase

• Largest Valley / max open drawdown dropped from $19,362 to $10,100
  That is a 47.8% reduction in max pain.

That last number is the standout.

The strategy did not just make more money per unit of risk.

It got dramatically easier to hold.

And that matters.

Because a strategy that looks good on paper but is psychologically brutal is not a strategy most traders will stick with.

Equity curve since 1997

Largest Valley / Max Pain

Results since 1997

What matters

The biggest improvement here was not just more profit.

It was much better return relative to drawdown and massively lower pain.

That is exactly the kind of improvement worth paying attention to.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

• This strategy is built for ES.D and shows institutional quality robustness because the same idea works across multiple, very different markets.

• Headline stats (ES.D test): 1.54 Profit Factor and 2528% Return / Max Drawdown.

• The “why” is simple: it waits for real intraday momentum, sizes everything to volatility, and avoids the low-quality chop where most breakout systems die.

• Paid section: plain-English rules + full code.

If a signal works across markets, it’s usually the signal not luck.

1) What problem this strategy solves

Most traders blow up breakouts for one reason:

They take every breakout.

This system is more selective. It tries to catch the days where the market is actually ready to move, not just wiggle.

We do that by combining:

• a simple “is today a good day to try?” filter (higher timeframe context), and

• a “price must prove it” entry (confirmation, not prediction).

We’re not forecasting. We’re filtering for the right days, then demanding proof.

2) Why this works particularly well on ES.D

ES is a great home for systematic intraday trading because it’s:

• deep and liquid (cleaner fills, less random spikes)

• structurally mean-reverting until it isn’t (when it trends, it can trend hard intraday)

So the edge comes from knowing when not to trade, and when to press.

This strategy tends to behave well on ES.D because it’s built around three “robust” ideas:

A) Volatility-scaled decisions

Instead of fixed tick levels, it scales:

• where we enter

• where we cut losses

• how we trail winners
  …based on current volatility.

That matters because ES doesn’t move the same way every month. Volatility expands and contracts. A fixed stop/entry is fragile. Volatility scaling is adaptive.

B) Confirmation entries

We don’t buy because “it looks strong.” We enter only after price pushes far enough to suggest the move is real. That avoids a lot of the chop that kills breakout systems.

C) Simple, mechanical exits

A protective stop + trailing stop gives a clean payoff profile:

• small controlled losses

• occasional large wins
  That asymmetry is what makes a breakout system worth running.

ES.D rewards systems that avoid noise, adapt to volatility, and only enter when momentum is real.

3) The real robustness test: it works across multiple markets

Here’s the part that matters more than any single metric:

We ran the same “signal DNA” across:

• ES.D

• RTY.D

• YM.D

• EMD.D

• NQ.D

• NK (day-session template only, because NK doesn’t have a .D equivalent)

These markets don’t just have different tick sizes, they have different drivers and different intraday personalities.

So if the strategy behaves consistently across them, that’s strong evidence we’re trading something structural (how markets move), not something local (one contract’s quirks).

NQ.D since 2001:

YM.D since 2002:

EMD.D since 2002:

NK since 2006 (Day Session Template):

Multi-market success is the closest thing to a “lie detector” for strategy robustness.

4) Results

ES.D Results since 1997

Profit Factor (PF) = 1.54

Profit Factor = total profits ÷ total losses

So 1.54 PF means:

• for every $1 the strategy lost (in total losing trades),

• it made $1.54 (in total winning trades).

Not perfect. Not magic. Just positive expectancy.

PF tells you if the system’s wins outweigh its losses overall.

Return / Max Drawdown = 2528%

This metric answers: “How much did it make compared to how much it hurt?”

Return / Max Drawdown = total return ÷ worst drawdown

So 2528% means the return was about 25.28x the worst peak-to-trough drawdown.

Why this matters:

• A strategy can have a good return and still be untradable if drawdowns are brutal.

• This ratio rewards strategies that grow without needing massive pain.

Return/MaxDD is “bang for your pain.” 2528% is exceptionally high.

5) Correlation: why this matters for a portfolio

If you’re building a real portfolio, the question isn’t:

“Do these strategies all make money?”

It’s:

“Do they lose money at the same time?”

This strategy is also designed as part of the M15.D series specifically to play well in a portfolio context especially alongside our earlier NQ work (notably: low loss correlation).

Diversification is about different drawdowns, not different tickers.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

Using a Higher Timeframe filter improved my baseline RSI strategy by 21% Ret/DD simply by conditioning entries on the daily structure. I did not touch the core signal. I only changed when I allow it to fire. Code below. Validate on your own data.

Subscribe now

If you trade indices like ES, NQ or RTY, this is one of the most robust filters you can add.

And here’s why.

The Problem With Single Timeframe Logic

Most traders build everything on one timeframe.

Entry. Exit. Stop. All on the same chart.

But markets are fractal. Intraday noise behaves very differently depending on what the higher timeframe is doing. A 15 minute pullback inside a strong daily uptrend is not the same as a 15 minute pullback inside a collapsing daily structure.

Same signal. Different context. Very different expectancy.

The goal is simple:

Do not change the edge.
Deploy it when the higher timeframe supports it.

Why Higher Timeframe Filters Are So Robust

This is one of the highest forms of robustness because:

• The higher timeframe is structurally independent from your entry timeframe
• It reduces overfitting by adding macro context
• It improves state conditioning without curve fitting entries

You are not optimising entries.
You are gating participation.

That distinction matters.

Higher timeframe filters tend to:

• Reduce trades during structurally weak environments
• Improve average trade
• Smooth drawdowns
• Lift Ret/DD without exploding parameter complexity

And unlike micro tweaks, this logic travels well across markets.

The Baseline Strategy

We use the same baseline RSI mean reversion strategy:

Long when RSI < 20 and price above 190 SMA
Exit when RSI > 80
Time exit and fixed stop

Here is the equity curve of the baseline:

Baseline Results:

Ret/DD: 6.24
Profit Factor: 1.53
Avg Trade: 279

Solid. Clean. But exposed to structural regime shifts.

The Higher Timeframe Filter

We now add a daily filter using Data2 (1440 minute bars of the same instrument).

Instead of just one condition, I built a Smart HTF Filter Switch.

You select a filter number from 1 to 14.
You scale its intensity using a multiplier from 1 to 12.

That gives you 168 combinations to test cleanly.

The filter options include:

• Daily SMA direction
• Close position within daily range
• Daily range expansion control
• Higher high and higher low structure
• Daily ADX trend strength
• Daily RSI zone alignment
• Bollinger width control
• Keltner position filter
• MACD alignment
• Stochastic structure
• ATR percentile regime
• Moving average ribbon stacking
• Rate of change momentum
• Pivot structure buffer

Each filter conditions entries on daily structure.

The baseline signal becomes:

LongEntry = BaseSignal AND FilterPass

That’s it.

No signal rewrite. No parameter distortion.

Code — Baseline Strategy (Reference)

//------------------------------------------------------------------
//Strategy: Long when RSI < 20 and price > 190 SMA, Exit when RSI > 80
//------------------------------------------------------------------

inputs:
	RSIPeriod( 3 ),
	RSILowEntry( 20 ),
	RSIHighExit( 80 ),
	MovingAvgPeriod( 190 ),
	ExitAfterBars( 39 ),
	StopLossTicks( 1100 ),
	PositionSize( 1 );

vars:
	RSIValue( 0 ),
	MovingAvg( 0 ),
	LongEntry( false ),
	LongExit( false );

// Calculate indicators
RSIValue = RSI(Close, RSIPeriod);
MovingAvg = Average(Close, MovingAvgPeriod);

// Entry conditions
LongEntry = (RSIValue < RSILowEntry) and (Close > MovingAvg[1]);

// Exit conditions  
LongExit = (RSIValue > RSIHighExit);

// Long Entry
if LongEntry and MarketPosition = 0 then begin
	Buy("RSI Long") PositionSize shares next bar at market;
end;

// Position Management
if MarketPosition > 0 then begin
	
	// Stop Loss
	if StopLossTicks > 0 then
		Sell("Stop Loss") next bar at (EntryPrice - StopLossTicks * MinMove/PriceScale) stop;
	
	// Exit after X bars
	if ExitAfterBars > 0 and BarsSinceEntry > 0 and BarsSinceEntry >= ExitAfterBars then
		Sell("Time Exit") next bar at market;
		
	// RSI Exit Signal
	if LongExit then
		Sell("RSI Exit") next bar at market;
		
end;

Code — Strategy With Higher Timeframe Filter

//------------------------------------------------------------------
// Strategy: RSI3 Long with Smart HTF Filter Switch
// Data2 = 1440-minute (daily) bars of the same instrument
//------------------------------------------------------------------
// Base: Long when RSI < 20 and price > 190 SMA, Exit when RSI > 80
//
// OPTIMIZE THESE 2 INPUTS:
//   FilterSelect  0-14  (0 = no filter, 1-14 = filter)
//   FilterMult    1-12  (scales the key parameter per filter)
//
// That gives you 168 combinations to test in the optimizer.
//------------------------------------------------------------------
//
// FILTER & MULTIPLIER REFERENCE
// ================================================================
// Sel | Filter                  | Mult Scales         | Formula            | Range
// ----|-------------------------|---------------------|--------------------|--------
//  0  | No filter               | -                   | -                  | -
//  1  | SMA Trend Direction     | SMA Length          | Mult * 10          | 10-120
//  2  | Close Position in Range | Bullish Min %       | 35 + Mult * 5      | 40-95
//  3  | Daily Range Size        | Lookback Length      | Mult * 5           | 5-60
//  4  | Higher High / Higher Low| Bars Back            | Mult               | 1-12
//  5  | ADX Trend Strength      | Trend Minimum        | 10 + Mult * 3      | 13-46
//  6  | RSI Zone                | Bullish Min          | 30 + Mult * 5      | 35-90
//  7  | Bollinger Band Width    | Max Width            | Mult * 0.01        | 0.01-0.12
//  8  | Keltner Channel         | Channel Multiplier   | Mult * 0.25        | 0.25-3.0
//  9  | MACD Signal Alignment   | Signal Length         | Mult + 2           | 3-14
// 10  | Stochastic Zone         | Overbought Level     | 60 + Mult * 3      | 63-96
// 11  | ATR Percentile Rank     | High Vol Ceiling     | 45 + Mult * 4      | 49-93
// 12  | Moving Average Ribbon   | Slow MA Length        | Mult * 10          | 10-120
// 13  | Rate of Change          | ROC Length            | Mult * 5           | 5-60
// 14  | Pivot Point Zone        | Buffer Strength      | Mult * 0.05        | 5-60% of range
// ================================================================

inputs:
    // === BASE STRATEGY ===
    RSIPeriod(3),
    RSILowEntry(20),
    RSIHighExit(80),
    MovingAvgPeriod(190),
    ExitAfterBars(39),
    StopLossTicks(1100),
    PositionSize(1),

    // === SMART FILTER SWITCH ===
    FilterSelect(0),    // 0=off, 1-14=filter number
    FilterMult(6);      // 1-12 parameter multiplier

vars:
    // Base strategy
    RSIValue(0),
    MovingAvg(0),
    LongEntry(false),
    LongExit(false),

    // Filter engine
    FilterPass(true),
    HTF_A(0),
    HTF_B(0),
    HTF_C(0),
    HTF_D(0),
    intrabarpersist HTF_i(0);


// ================================================================
// FILTER ENGINE
// ================================================================
FilterPass = true;

if FilterSelect = 1 then begin
    // ---------------------------------------------------------
    // SMA Trend Direction
    // SMA Length = FilterMult * 10 (range: 10 to 120)
    // Pass when daily SMA is rising
    // ---------------------------------------------------------
    HTF_A = Average(Close of Data2, FilterMult * 10);
    FilterPass = HTF_A > HTF_A[1];
end

else if FilterSelect = 2 then begin
    // ---------------------------------------------------------
    // Close Position in Range
    // BullMin = 35 + FilterMult * 5 (range: 40 to 95)
    // Pass when daily close is in upper portion of range
    // ---------------------------------------------------------
    HTF_A = High of Data2 - Low of Data2;
    if HTF_A > 0 then
        HTF_B = ((Close of Data2 - Low of Data2) / HTF_A) * 100
    else
        HTF_B = 50;
    FilterPass = HTF_B >= 35 + FilterMult * 5;
end

else if FilterSelect = 3 then begin
    // ---------------------------------------------------------
    // Daily Range Size
    // Lookback = FilterMult * 5 (range: 5 to 60)
    // Pass when today's range is 50-200% of average
    // Shorter lookback = more reactive, longer = more stable
    // ---------------------------------------------------------
    HTF_A = High of Data2 - Low of Data2;
    HTF_B = Average(High of Data2 - Low of Data2, FilterMult * 5);
    if HTF_B > 0 then
        HTF_C = (HTF_A / HTF_B) * 100
    else
        HTF_C = 100;
    FilterPass = HTF_C >= 50 and HTF_C <= 200;
end

else if FilterSelect = 4 then begin
    // ---------------------------------------------------------
    // Higher High / Higher Low
    // Bars Back = FilterMult (range: 1 to 12)
    // Pass when today's high/low exceed N days ago
    // Higher mult = requires trend over longer period
    // ---------------------------------------------------------
    FilterPass = High of Data2 > High[FilterMult] of Data2 and
                 Low of Data2 > Low[FilterMult] of Data2;
end

else if FilterSelect = 5 then begin
    // ---------------------------------------------------------
    // ADX Trend Strength
    // Trend Min = 10 + FilterMult * 3 (range: 13 to 46)
    // Pass when daily ADX shows sufficient trend strength
    // Low mult = weak trend ok, high mult = strong trend required
    // ---------------------------------------------------------
    HTF_A = ADX(14) of Data2;
    FilterPass = HTF_A >= 10 + FilterMult * 3;
end

else if FilterSelect = 6 then begin
    // ---------------------------------------------------------
    // RSI Zone
    // Bull Min = 30 + FilterMult * 5 (range: 35 to 90)
    // Pass when daily RSI shows bullish momentum
    // Low mult = mild bullish ok, high mult = strong required
    // ---------------------------------------------------------
    HTF_A = RSI(Close of Data2, 14);
    FilterPass = HTF_A >= 30 + FilterMult * 5;
end

else if FilterSelect = 7 then begin
    // ---------------------------------------------------------
    // Bollinger Band Width
    // Max Width = FilterMult * 0.01 (range: 0.01 to 0.12)
    // Pass when bands are NOT too wide (overextended)
    // Low mult = very selective, high mult = more permissive
    // ---------------------------------------------------------
    HTF_A = Average(Close of Data2, 20);
    HTF_B = HTF_A + 2 * StdDev(Close of Data2, 20);
    HTF_C = HTF_A - 2 * StdDev(Close of Data2, 20);
    if HTF_A > 0 then
        HTF_D = (HTF_B - HTF_C) / HTF_A
    else
        HTF_D = 0;
    FilterPass = HTF_D < FilterMult * 0.01;
end

else if FilterSelect = 8 then begin
    // ---------------------------------------------------------
    // Keltner Channel Position
    // Channel Mult = FilterMult * 0.25 (range: 0.25 to 3.0)
    // Pass when price is above the lower Keltner band
    // Low mult = tight channel (easy pass), high = wide (easy pass)
    // Reducer: very low mult makes a tight channel that filters
    //          more aggressively during selloffs
    // ---------------------------------------------------------
    HTF_A = Average(Close of Data2, 20);
    HTF_B = HTF_A - FilterMult * 0.25 * AvgTrueRange(10) of Data2;
    FilterPass = Close of Data2 >= HTF_B;
end

else if FilterSelect = 9 then begin
    // ---------------------------------------------------------
    // MACD Signal Alignment
    // Signal Length = FilterMult + 2 (range: 3 to 14)
    // Pass when MACD line is above signal line
    // Low mult = fast signal (more whipsaw), high = slow (smoother)
    // ---------------------------------------------------------
    HTF_A = MACD(Close of Data2, 12, 26);
    HTF_B = XAverage(HTF_A, FilterMult + 2);
    FilterPass = HTF_A > HTF_B;
end

else if FilterSelect = 10 then begin
    // ---------------------------------------------------------
    // Stochastic Zone
    // OB Level = 60 + FilterMult * 3 (range: 63 to 96)
    // Pass when %K > %D and %K is not overbought
    // Low mult = strict OB ceiling, high mult = permissive
    // ---------------------------------------------------------
    HTF_A = SlowK(14) of Data2;
    HTF_B = SlowD(3) of Data2;
    FilterPass = HTF_A > HTF_B and HTF_A < 60 + FilterMult * 3;
end

else if FilterSelect = 11 then begin
    // ---------------------------------------------------------
    // ATR Percentile Rank
    // High Vol Ceiling = 45 + FilterMult * 4 (range: 49 to 93)
    // Pass when volatility is NOT in the extreme upper percentile
    // Low mult = strict (skip moderate+ vol), high = permissive
    // ---------------------------------------------------------
    HTF_A = AvgTrueRange(14) of Data2;
    HTF_B = 0;
    for HTF_i = 1 to 100 begin
        if HTF_A[HTF_i] < HTF_A then
            HTF_B = HTF_B + 1;
    end;
    HTF_C = (HTF_B / 100) * 100;
    FilterPass = HTF_C < 45 + FilterMult * 4;
end

else if FilterSelect = 12 then begin
    // ---------------------------------------------------------
    // Moving Average Ribbon
    // Slow MA Length = FilterMult * 10 (range: 10 to 120)
    // Fast = 10, Med = 20, Slow = scaled by mult
    // Pass when fast > med > slow (full bullish stack)
    // Low mult = tight ribbon (responsive), high = wide (lagging)
    // ---------------------------------------------------------
    HTF_A = Average(Close of Data2, 10);
    HTF_B = Average(Close of Data2, 20);
    HTF_C = Average(Close of Data2, FilterMult * 10);
    FilterPass = HTF_A > HTF_B and HTF_B > HTF_C;
end

else if FilterSelect = 13 then begin
    // ---------------------------------------------------------
    // Rate of Change Momentum
    // ROC Length = FilterMult * 5 (range: 5 to 60)
    // Pass when daily ROC is positive (upward momentum)
    // Low mult = short-term momentum, high = long-term trend
    // ---------------------------------------------------------
    if Close[FilterMult * 5] of Data2 > 0 then
        HTF_A = ((Close of Data2 - Close[FilterMult * 5] of Data2) /
                  Close[FilterMult * 5] of Data2) * 100
    else
        HTF_A = 0;
    FilterPass = HTF_A > 0;
end

else if FilterSelect = 14 then begin
    // ---------------------------------------------------------
    // Pivot Point Zone
    // Buffer = FilterMult * 0.05 of yesterday's range (5-60%)
    // Pass when price is above pivot + buffer
    // Low mult = just above pivot ok, high = well above required
    // Uses standard pivot formula from prior day HLC
    // ---------------------------------------------------------
    HTF_A = (High[1] of Data2 + Low[1] of Data2 + Close[1] of Data2) / 3;
    HTF_B = (High[1] of Data2 - Low[1] of Data2) * FilterMult * 0.05;
    FilterPass = Close of Data2 > HTF_A + HTF_B;
end;


// ================================================================
// BASE STRATEGY
// ================================================================

RSIValue = RSI(Close, RSIPeriod);
MovingAvg = Average(Close, MovingAvgPeriod);

LongEntry = (RSIValue < RSILowEntry) and (Close > MovingAvg[1]);
LongExit = (RSIValue > RSIHighExit);

if LongEntry and FilterPass and MarketPosition = 0 then begin
    Buy("RSI Long") PositionSize shares next bar at market;
end;

if MarketPosition > 0 then begin

    if StopLossTicks > 0 then
        Sell("Stop Loss") next bar at (EntryPrice - StopLossTicks * MinMove/PriceScale) stop;

    if ExitAfterBars > 0 and BarsSinceEntry > 0 and BarsSinceEntry >= ExitAfterBars then
        Sell("Time Exit") next bar at market;

    if LongExit then
        Sell("RSI Exit") next bar at market;

end;

Requires Data2 = 1440 minute bars of same instrument.

Results With Higher Timeframe Filter

After optimisation over:

FilterSelect: 0–14
FilterMult: 1–12

Best performing configuration:

FilterSelect = 4
FilterMult = 8

Improvement:

Ret/DD: 6.24 → 7.57
Lift: 21%
Profit Factor: 1.53 → 1.851.85
Avg Trade: 279→ 376

Here is the new equity curve:

Notice:

• Fewer low quality trades
• Shallower drawdown clusters
• Smoother equity slope

Why This Works

Higher timeframe filters:

1. Remove structurally weak participation windows

2. Improve state conditioning

3. Reduce drawdown clustering

4. Travel better across instruments

They are robust because they operate at a structural level, not a signal tweak level.

You are not trying to predict more.
You are trying to participate smarter.

Disclaimer

Educational only. Not financial advice. Results include hypothetical backtests. Futures involve substantial risk. Validate on your own data and infrastructure before deploying capital.

It’s like Rocky Balboa in the 15th round no matter how hard the market hits it, it just refuses to stay down. This is one of those rare cases. Today, I’m sharing a pure day trading model that has me genuinely excited.

The “Twist”: Adaptive Bias Currently, the model is leaning heavily into shorts. While the entry logic is perfectly symmetrical, the prevailing market regime is rewarding the downside. This is exactly what a rules based system should do: respond to actual conditions rather than predicting them.

Proven Robustness What gives me confidence is the “portability” of the edge. I applied the exact same rules with zero tweaks to ES.D, YM.D, EMD.D RTY, and even Nikkei (NK) using a U.S. session template. Because these symbols act as “unseen data,” the fact that the performance held proves this isn’t a curve fitted mirage. The signal is fundamentally strong enough to transcend a single asset class.

The Framework:

• Instrument: NQ.D (Day Session)

• Primary Chart: 15-minute bars (Data1)

• Trend Filter: 1440-minute daily series (Data2)

• Execution: Max one trade per day; flat by the close.

• Risk: Volatility-aware management.

Results

Optimised for the NQ.D (Day Session), this strategy delivered a solid 1.59 Profit Factor and a 46% win rate. Most impressively, it achieved a remarkable 24x Return on Max Drawdown (ROMAD). These results were produced using strictly fixed 1-contract position sizing with no compounding involved.

Raw Metrics: Cross-Symbol Validation

ES.D since 1997

EMD.D since 2002

YM.D since 2002

RTY since 2001

NK since 2006 (US Day Session)

🧪 Robustness testing

I trust results only after they pass the usual Stress tests.

Monte Carlo parameters

We vary the parameters to test stability; if performance holds, it’s a green flag that the strategy isn’t reliant on precise settings. (The tight bands shows us the strategy is very stable and adjusting the parameters will still give us a positive result)

Monte Carlo extreme parameter stress test

In this test we are more looking to see what happens if we stretch the parameters to see if the strategy falls apart.

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more

📌 In case you missed the recent strategies

🧭 Why a portfolio (for intraday)?

Single systems have personality: time of day bias, volatility preference, and streaky PnL. A portfolio smooths variance by mixing uncorrelated edges and reduces dependence on any one time segment or symbol.

Key goals:

• Lower correlation between components

• Avoid overlapping weak windows

• Keep every rule simple and interpretable

📈 Portfolio: combined results (3 strategies)

Last 2 years: This portfolio did a unbelievable +154% on a 50k starting balance, which works out to be ~74% return per year, fixed size (no compounding).

Equity of the last 2 years:

Since 2002: ~40%/yr. Note: because position size is fixed, CAGR looks lower than a compounded run.

🔗 Overlap & correlation (don’t double up the same risk)

We measure daily and weekly PnL correlations to keep systems that win and lose at different times.

What we’re looking for

• Low-to-moderate correlations (not all the same sign)

• No single system dominating trades or risk

Daily correlation matrix:

🧪 Stress & robustness (portfolio level)

We run Monte Carlo at the portfolio level to get a more realistic distribution of outcomes.

What we test

Randomise & skip trades (execution variance, outages)

✅ Takeaway

This portfolio works because it spreads risk across simple, complementary ideas that don’t move together. The correlation checks back that up across daily and weekly views.
Monte Carlo shows the curve remains resilient through tougher sequences, and fixed-size testing avoids inflated claims. The core message is steady progress with controlled drawdowns.

⚠️ Disclaimer

This publication is education only, not financial advice. Results include hypothetical backtests; past performance is not indicative of future results. Futures and derivatives carry substantial risk (leverage, gaps, slippage, liquidity, data/execution issues). Validate on your own data, costs, and infrastructure before using real capital.

This week we’ve got a 30-minute breakout strategy for Gold. And no, this isn’t “trade the hot market of the month.” Please don’t run a system just because an asset is trending on social feeds. Think in portfolios. As Ray Dalio puts it, diversification is the closest thing to a free lunch.

Gold earns its seat in a portfolio because it often behaves differently to equities and bonds, especially in uncertainty. That’s exactly what we’re seeing now.

We keep it simple. Over complicated systems look great in a backtest and then fall apart live. We’d rather run simple, robust ideas across many markets than bet on a single hero system. Robustness first; the portfolio does the heavy lifting.

📖 The Idea

The idea is simple: only trade when strength is there, make price prove it with a breakout, cap risk with an ATR stop, and exit when strength fades.

One entry, one stop, one exit. It runs long or short with mirror rules and flattens at 15:00.

🧠 How it works in plain English

Keep it simple. One entry, one stop, one exit. It works long and short with mirror rules.

We check trend strength. When strength permits, we place a stop order at a breakout level.

Price must trade through that level for us to enter.

Risk is capped with a ATR stop that adapts to volatility.

We exit on a open based average once the original strength condition fades.

All positions are flattened at 15:00 exchange time, and new orders can arm and trigger any time before that cut off, including overnight.

Our simulations consistently show that excluding overnight entries negatively impacts performance. While the data favors a 24-hour approach, you may still opt to restrict entries to the day session based on your preference

🎯 Why this fits GC

Gold spends most of its time coiled until it explodes. Since the biggest moves are around macro data, we use breakout entries to ensure we only catch the real velocity.

A trend filter acts as a noise canceller, keeping you sidelined during the chop and only “arming” the system when there’s genuine momentum.

On GC, ATR-based sizing isn’t optional; it’s a necessity. It allows the stop to breathe during high volatility spikes and contract when the market settles.

To handle Gold’s habit of aggressive runs followed by immediate reversals, a smoothed exit captures profit the moment strength weakens.

We flatten the book at 15:00 to lock in the day’s liquidity and dodge the “dead zone” drift and overnight gap risk.

Whether it’s ripping on inflation or dumping on a strong dollar, the logic is perfectly symmetric for both sides.

📈 Results of strategy since 2002

Equity of strategy (30-minute bars)

I also want to add that 2002 to 2007 was left completely untouched (true Out Of Sample). We ran the model once and did not use that window in development. That is about as close to live as you can get, and the equity over that stretch speaks for itself.

🧪 Robustness testing

I trust results only after they pass the usual Stress tests.

Monte Carlo parameters

We vary the parameters to test stability; if performance holds, it’s a green flag that the strategy isn’t reliant on precise settings. (The tight bands shows us the strategy is very stable and adjusting the parameters will still give us a positive result)

Monte Carlo extreme parameter stress test

In this test we are more looking to see what happens if we stretch the parameters to see if the strategy falls apart.

Monte Carlo randomise and skip trades

We shuffle the order of historical trades and randomly remove some to mimic misses, slippage, or bad fills. If performance stays healthy, the edge is not just luck in trade order. Even at 95% we still have a healthy return to max drawdown.

Walk Forward Matrix

We test the strategy by dividing historical data into rolling periods, optimising parameters on one period, then testing on the next unseen period. When results stay consistent across multiple windows, it proves the strategy adapts to changing markets rather than being curve-fit to past data.

System Parameter Permutation test

We run many nearby parameter combinations around the chosen parameters. If most versions work, the model is stable and not tuned to one narrow sweet spot. As you can see our net profit is within the range of our median which shows that we havent choosen lucky parameters for our strategy.

Results on alternative market (To check if robust or overfit) run on SI (Silver)

We apply the same model on SI futures without adjusting the model (raw out of sample). If the results are acceptable, the idea is general and not tied only to GC.

Results and Equity with Slippage and Commision + Inside bar back testing

Ready to put it to work? The full EasyLanguage code is below.

Dont forget this code can be easily converted to any other trading platform using LLMs and of course this can be run on Micro aswell.

By request, I’ll now include a clear plain English breakdown of each strategy with every parameter so you can rebuild it on your platform.

Read more