I Mastered Chaos Theory to Develop a 3.275 Sharpe FX Strategy. Here's How to Do It.
Empirical evidence around how markets follow fractal like behavior leads us to highly profitable and regime resistant currency trading strategies.
Chaos theory is a fancy topic that alludes to a world almost as mystic and mysterious as quantum physics. Everyone has heard of it, but not many people really know what it means or how to use it.
Fortunately for us, you can’t judge a book by its cover. And the reality is that chaos theory is simply a way for us to describe behavior in what otherwise looks like complete randomness (aka the markets).
By harnessing the power of these concepts, I was able to put together a highly profitable, low risk strategy in the foreign exchange markets that reached a Sharpe ratio of over 3.
Here’s how I did it:
Rather than shying away from the preconceptions, I dove into deep and found a mountain of empirical evidence and research that proved the water wasn’t so bad after all.
And you should, too. What are you avoiding based on your negative assumptions that could be holding you back from the life you want to live?
The Perimeter of Great Britain is Infinity (really)
Such a premise seems ridiculous. But, this is the classic Coastline Paradox. If we take stakes and place one every 10 kilometers around the coastline of Great Britain and count them up, it will measure less than if we place one every 5 kilometers…
…which is less than if we place them every 2.5 kilometers, which is less than…
You get the point.
In fact, if you could place an infinite number of stakes around the coast of Great Britain, it would measure infinite kilometers long.
But how? Because the perimeter is not definite, but measured as a fractal: the closer you look, the more detail you’re able to see. And just like a coastline, the market behaves in the same way. We can see more detail and a totally different picture if we zoom into pure ticks versus if we look at the daily charts.
The Measuring Stick of Market Movement - Theta Θ
Research out of the University of Essex back in 2018 actually formalized this concept in an application to the markets. They proposed not looking at the markets on the basis of time, like we normally do, but to look at it in terms of distance or fixed price movements.
So rather than choosing to place stakes every 10 kilometers or 5 kilometers, researchers looked at the market placing points at every 0.1% movement or 0.2% movement, for example.
This variable, which is like a market ruler, was then explored with the curiosity of wanting to see what patterns emerged from a different lens. What they found is that the market, when that ruler is properly tuned, actually performs rather predictively.
Researchers called this variable length of the ruler theta Θ and below is looking at the market in terms of directional changes or pivot points when Θ is set to certain values.
This study was to measure if there was directional continuity between Θ.
So, if the market moved up one jump, would it move up another amount of Θ next? If the market moved up 1 unit of Θ, would the next jump be 0.5 units of Θ?
Yes. There exist values of Θ for certain markets that can describe actual momentum and directional continuity which can be exploited into trading strategies.
Overshooting the Market Momentum - Finding Corrections
So, if we know that after 2 units up of Θ, the market will continue for 1 unit up of Θ, what happens after?
Well, you would expect the market to correct itself, which is exactly what it does in certain scenarios. If the market moves too far outside of its expected behavior, it will snap back in a correction, giving us the perfect opportunity to strike.
Look more carefully at the chart above. After the directional change event (DC), the market enters an overshoot event (OS) and immediately after snaps back to the ‘correct’ price.
Specifically, 0.2% of negative price movement, followed by another 0.2% of negative price movement, lead to a reversal of 0.2% of positive price movement.
In a nice simple chart, you can think of the market as a state machine of three phases. If it passes the first two, then you can bet the third is going to happen.
Great. But a paper is a paper. Let’s test this on foreign exchange data to see if it actually holds up.
Making the Trade
I downloaded 10 years of 1-minute FX data from EODHD and implemented the algorithm in the paper I’ve been referencing (available here via open access).
The first step is to find the optimal value of Θ in which we are maximizing the amount of profit we are making via the 3-phase strategy above. To do this, we do a simple split window system that does a grid search through a sweep of Θ values - landing on the ones that is most profitable and least risky.
The paper also includes a retraining catch which will re-optimize for Θ if the out-of-sample drawdown exceeds the in-sample drawdown.
It also includes a small decision tree that looks at the standard deviations of prices along with the time each overshoot to the beginning of the reversal to create a binary classifier for trades in-sample.
In other words, given a trade’s Phase 2 length and the standard deviation of prices in that time, predict if that trade was profitable or not.
While sounding complex, the implementation was nothing crazy. It was a simple parameter sweep to determine the best Θ and downward overshoot indicator values, and then train a decision tree on the trades that popped out based on that.
I took 8 symbols and exploded them out into 36 tradable symbols by creating synthetic forex legs such that:
Which trained from 2024 → 2025 and tested from 2025 → mid-2026 (the time of writing) produced some pretty good results given that the paper was written 8 years ago.
Making it Meta - A Dynamic Basket Strategy
It wouldn’t be a Paper to Profit post without some sort of modification and remixing. And so, I added a simple rolling Sharpe ratio function that looks at all the 30 valid (6 symbols contained non-numeric and invalid results) out-of-sample equity curves to develop a well diversified portfolio strategy.
The metastrategy outperforms equal-weighting and dynamically allocates between the market and cash thanks to our position sizing algorithm, which means that we can use the rest of the cash for another system we have running as an additional hedge or stream of revenue.
A Chaotic Conclusion
And so, in conclusion, chaos theory actually isn’t that hard to understand, nor that confusing. It also describes the market in pretty simple terms (when we apply one aspect of chaos theory to the market as there are hundreds of different concepts in the entire school of thought).
But now that you can look at the market through a different lens, how else can you profit? Can this be applied to stop loss systems? Does it port over to other markets? Does using log-returns instead of simple returns improve performance?
All these questions, and more, now at your fingertips, especially if you’re a paid subscriber because you can download the notebook and run it to recreate this exact strategy today via the Google Drive:
Not Yet a Paid Subscriber?
How silly of you. Become one now and get access to all the code I used to make this blog post and code from my previous posts!
All the best, and happy trading!
Interesting approach with chaos theory for FX. We found that regime classification is the foundation for any strategy resilience. A 3.27 Sharpe in FX is impressive, but the real test is how it performs across regime shifts, not just in-sample. Do you track performance separately across vol regimes? That is where most strategies break
a 3.275 sharpe on fx is the part that stops me. that precision is usually where the overfit hides. the fractal stuff i buy in spirit but every regime resistant claim i've traded held right up until the regime i hadn't seen yet. what's the out of sample window look like, and does the sharpe survive costs?