Mean reversion is the mirror image of breakout trading. Where a breakout system buys strength, a mean reversion system sells it, betting that an overextended move will snap back toward its average. Both work. They just work at different times, which is exactly why we run them together.
The core idea
Currencies, unlike stocks, do not trend to zero or infinity. A currency pair is a ratio between two economies, and ratios oscillate. When price stretches too far too fast from its recent average, especially during quiet liquidity windows without news catalysts, the probability of a pullback rises. A mean reversion system quantifies too far and too fast, then fades the move with a defined invalidation point.
The win rate profile
Our mean reversion system runs at a 72% win rate across 1,438 backtested trades. That is the opposite profile of the breakout system (57% win rate, bigger winners). Mean reversion collects many small wins and occasionally takes a larger loss when a move that looked overextended just keeps going. This is why the system trades range-bound pairs like AUDUSD, EURCHF, CADCHF, USDCAD, AUDCHF, and GBPAUD rather than trending monsters like GBPJPY.
Why the two systems belong together
Here is the part most EA buyers miss. In our combined backtest, something counterintuitive happens: if you lower the risk on the mean reversion system while keeping breakout risk constant, the total portfolio drawdown gets worse, not better. In December 2017, the breakout system had its worst month while mean reversion was positive. The mean reversion profits absorbed part of the breakout losses. Cut mean reversion in half and you cut the shock absorber in half.
This is diversification in its purest form: two systems with different logic, different pairs, and negatively correlated worst months. The combined portfolio has had zero losing years since 2015 in backtest, something neither system fully guarantees alone.
The honest downsides
- Mean reversion feels bad to trade manually. You are selling into strength, which is psychologically brutal. Automation removes that problem.
- The occasional large loss is real. A 72% win rate means the 28% must stay controlled, which is what the fixed invalidation rules are for.
- It underperforms in strong trend years, which is when the breakout side of the portfolio carries the load.
The Mean Reversion System is available standalone, but the data argues strongly for running it inside The Full System.
The anatomy of a typical trade
A concrete example on AUDUSD. Over two quiet sessions without economic releases, the pair grinds 85 pips above its recent equilibrium, not on news but on flow: a large order being worked, stops being cleaned out, nothing fundamental. The system's overextension metrics cross their thresholds, and a short entry triggers with a stop placed beyond the recent extreme, sized to the configured risk. The trade thesis is not that AUDUSD is going down. It is that this specific move is stretched beyond what typically sustains without a catalyst, and the path of least resistance is a partial retracement.
Two sessions later, the pair has drifted 40 pips back toward equilibrium and the system exits at its target. Win number 1,042 of the backtest's 1,438 trades, entirely unremarkable, which is the aesthetic of mean reversion: no home runs, just a high percentage of singles. The losing version of this trade happens when a genuine catalyst emerges mid-position and the stretch becomes a trend. The stop takes the loss, larger than the average win, and the 72% win rate absorbs it over time.
Why we exclude JPY pairs from mean reversion
Notice the pair list: AUDUSD, GBPUSD, EURCHF, CADCHF, USDCAD, AUDCHF, GBPAUD. No yen. This is deliberate and it mirrors the reason the breakout system trades only yen. The same rate-differential and risk-sentiment forces that make yen trends persist (a gift to breakout logic) make yen overextensions dangerous to fade: what looks stretched keeps stretching because a structural driver is behind it. The mean reversion pairs are chosen for the opposite property: cross rates between economies with correlated monetary conditions, where large dislocations lack persistent fuel and equilibrium pull is strong. Strategy-instrument fit is not a detail. It is half the edge, and it is why running both systems means trading ten different pairs with zero overlap.