Stefan Friedrichowski has a background in physics, not finance. He has been trading for over 30 years, and for the last decade has done it full-time – applying the experimental thinking of his scientific training to build algorithmic mean reversion strategies for forex markets. He now develops strategies professionally as a freelancer for JFD Bank, with a fully systematic, no-discretionary approach using Expert Advisors.
His entry point into mean reversion was a single statistic he encountered online: that 70% of the time, markets are ranging rather than trending. That number, even if it was hard to prove precisely, shifted his entire research direction. If the market is not trending most of the time, why build only trend-following systems?
Watch the full episode below, then read on for the complete breakdown.
Why Renko-Style Bricks Reveal Mean Reversion Edges
Stefan’s first breakthrough was moving away from time-based candles and toward price-based bricks – similar to Renko charts but with a fixed percentage height rather than a fixed pip count. This percentage-based approach normalizes across currency pairs regardless of their absolute price levels, making patterns comparable between EUR/USD and AUD/CAD, for example.
When he plotted these bricks on AUD/CAD with a 0.3% brick height, a pattern emerged clearly: after a white (up) brick, a black (down) brick followed far more often than chance would predict, and vice versa. This alternating behavior is the most basic expression of mean reversion – the market reverting after each small move.
“After a white brick, we have a black brick in a lot of cases. That is already the first very good hint for an edge. We see something which is not random. That’s always the starting point of any trading strategy.”
The key discipline: Stefan confirmed this pattern not just for the example he showed, but across 15 years of data and across multiple currency pairs. The edge had to be robust over time and across markets before it was worth building a strategy around.
Identifying Markets Suited to Mean Reversion
Not all forex pairs behave the same way. Stefan developed a method to test which markets were genuinely mean-reverting versus trending before committing to building a strategy.
His process: take a sequence of bricks, count how often an up-brick is followed by a down-brick (and vice versa), and compare that frequency to what pure randomness would produce. If the alternating frequency is meaningfully higher than 50%, the market is exhibiting mean-reverting behavior. If up-bricks tend to be followed by more up-bricks, the market is trending.
This statistical test, run over many years of data, produced a ranking of currency pairs from most mean-reverting to most trending. Some pairs consistently showed reverting behavior. Others did not. That ranking directly determined which markets Stefan would build strategies for and which he would avoid.
The Role of Statistics and the Physics Mindset
Stefan is explicit that his advantage comes from treating strategy development like a scientific experiment. You form a hypothesis, design a test, run it across a large enough sample, and look at whether the result is replicable across different markets and time periods.
“Trading is like doing an experiment. You use what you know about physics, statistics, and experiment design. You come to rules, then you test them.”
This matters for mean reversion in particular because the patterns can look compelling on cherry-picked examples. Stefan’s discipline is to always run any pattern across the full 15-year data set across multiple pairs. If it only works on the single example he showed to illustrate it, he ignores it.
The sample size argument is central to his work. Time-based candles on short timeframes produce thin data. Price-based bricks at a 0.3% size produce far more bricks per year, giving him hundreds and thousands of observations to work with. That makes the statistics meaningful.
Building the Strategy Around the Edge
Once Stefan identifies a mean-reverting market with a demonstrated edge at the brick level, he builds entry and exit rules around it. The core logic: after a down-brick, look to go long (and vice versa). But the raw alternation signal alone is not enough to build a profitable strategy – the entry timing, stop loss placement, and profit target all need to be defined.
His approach to exits is systematic. He tests three options for each setup: a fixed profit target, a trailing stop, and a time-based exit. Each pattern type has a different optimal exit method, and he does not assume the same approach works for every setup.
Position sizing is equal and percentage-based across all pairs, ensuring no single trade dominates the portfolio. This is consistent with his physics background – equalize the variables before drawing conclusions about what drives performance.
Common Pitfalls in Mean Reversion Strategy Development
Stefan identified several mistakes that traders make when attempting mean reversion strategies on their own. The most common is testing on time-based candles, where the sample sizes are too small to draw statistical conclusions – particularly on longer timeframes like daily bars.
The second is testing on markets that are not actually mean-reverting. Running a mean reversion strategy on a strongly trending pair will produce losses by design. The market identification step – which Stefan runs before any strategy development – is a prerequisite, not an optional extra.
The third pitfall is over-optimizing. Because mean reversion entries tend to have good win rates (the market does revert most of the time), it is tempting to optimize parameters to push the win rate even higher. But over-optimized parameters fail out of sample. Stefan’s filter: does the strategy hold up across many different parameter values, and does it still work across different pairs?
Algorithmic Execution and the Full-Time Setup
Stefan’s strategies are fully algorithmic, running as Expert Advisors with no discretionary override. This is not just a preference – it is necessary for the strategy to work. Mean reversion setups often look bad on a chart. The entry is typically into a declining price, and the chart may look like the stock or currency pair is simply falling. Discretionary traders tend to cancel these entries. That destroys the edge.
“When you look at these charts, you often see a currency pair going straight down. One of the hardest parts is placing your order. I purposely don’t look at the charts because if you look, you ask why do I want to buy this? And often those are actually the best entries.”
Running the strategies without watching them is part of the methodology, not a convenience. The data does the work. The trader’s job is to build the strategy correctly and then get out of the way.
Why Forex Works Well for Mean Reversion
Stefan’s preference for forex over equities or futures comes down to symmetry. Forex allows short and long trades in both directions with equal ease and equal cost, which is essential for a mean reversion system that needs to trade both sides of the market to capture all the edge.
Equity markets have structural upward bias that complicates the short side. Futures require specific knowledge of contract roll, margin, and liquidity. Spot forex, traded with a reputable broker and percentage-based position sizing, provides the cleanest environment for the statistical testing Stefan’s approach relies on.
Related episodes
- Building Mean Reversion Trading Strategies with Cesar Alvarez – Part 1
- Low-Effort Trading Strategies with Cesar Alvarez
- Combining Algos Using State-Based Market Design with Richard Metzger
Want more on building mean reversion trading strategies? Subscribe to the Better System Trader podcast for weekly interviews with the world’s top systematic traders and quantitative researchers.
