Backtesting Forex Mean Reversion on Major Currency Pairs

# Backtesting Forex Mean Reversion on Major Currency Pairs

The foreign exchange market is the most liquid market in the world, trading over $7 trillion daily. This liquidity and the range-bound nature of many currency pairs make forex a potentially ideal market for mean reversion strategies. We backtested several approaches to determine where the edge lies.

Why Forex Might Mean-Revert

Currency pairs have fundamental anchors (purchasing power parity, interest rate differentials, trade balances) that prevent extreme persistent trends in most cases. Unlike stocks which can appreciate indefinitely, currencies exist in pairs where one side strengthening necessarily means the other weakening. Central bank interventions also tend to cap extreme moves. These structural features create conditions favorable for mean reversion.

Test Universe and Data

We tested on the seven major currency pairs: EUR/USD, GBP/USD, USD/JPY, USD/CHF, AUD/USD, USD/CAD, and NZD/USD. Data spans 2005-2024 using 4-hour and daily timeframes. We used bid/ask data to account for spread costs and assumed no additional commission (reflecting modern retail forex pricing). All tests used a fixed fractional position sizing of 1% risk per trade.

Strategy 1: RSI Mean Reversion

Using a 14-period RSI on daily data, we entered long when RSI dropped below 25 and short when RSI rose above 75. Exit occurred when RSI returned to the 50 level. Across all seven pairs, this generated 892 trades with a 51.3% win rate and a profit factor of 1.08. After spreads, the edge was negligible. The traditional RSI approach simply does not work well on daily forex data.

Strategy 2: Modified RSI with Trend Filter

We adapted the equity RSI approach for forex: 2-period RSI below 10 for entries, with a 100-period SMA as a trend filter (only trade in the trend direction). This dramatically improved results to a 58% win rate and 1.34 profit factor across 1,247 trades. EUR/USD and GBP/USD showed the strongest results. USD/JPY performed worst, likely due to its tendency for persistent carry-driven trends.

Strategy 3: Bollinger Band Reversion

We tested buying at the lower Bollinger Band (2 standard deviations) and selling at the upper band on 4-hour data. A key modification was requiring the band touch plus a reversal candle (close above open for longs). This produced a 56% win rate with 1.42 profit factor. The 4-hour timeframe significantly outperformed daily data for this approach, generating more signals with similar quality.

Strategy 4: Z-Score of Spread from Moving Average

This approach calculates how far price has deviated from a 50-period moving average, expressed in standard deviations. Enter mean reversion trades when the z-score exceeds 2.5, targeting a return to the 0 level. This was our best-performing strategy with a 61% win rate and 1.53 profit factor on 4-hour data. The z-score approach naturally adapts to changing volatility regimes, making it more robust than fixed-threshold methods.

Session Timing Effects

A significant finding was that mean reversion worked substantially better during the Asian trading session compared to London or New York. The Asian session (0:00-8:00 GMT) had win rates 7% higher than other sessions for mean reversion entries. This aligns with the observation that the Asian session is typically range-bound, while London and New York sessions tend to generate directional breakouts that punish mean reversion.

Carry Trade Interaction

Currency pairs with large interest rate differentials showed weaker mean reversion results. When the carry trade is active, high-yielding currencies trend persistently against low-yielders, frustrating mean reversion strategies. Our backtesting showed that filtering out trades that fight the carry direction improved the profit factor by 0.15 on affected pairs. Understanding the macro context improves mechanical strategy results.

Drawdown Analysis

The maximum drawdown across our best strategy was 11.3%, occurring during the Swiss Franc unpegging event in January 2015. Flash crashes and central bank surprises create tail risk that mean reversion strategies are particularly vulnerable to because they naturally position against extreme moves. A hard stop at 3x the normal stop distance provided protection against these events with minimal impact on overall performance.

Multi-Pair Portfolio Effects

Running the strategy simultaneously across all seven pairs provided significant diversification benefits. The portfolio Sharpe ratio was 1.8 compared to individual pair Sharpe ratios ranging from 0.6 to 1.2. Correlations between pair equity curves were low (averaging 0.15), meaning drawdowns in one pair rarely coincided with drawdowns in others. This diversification effect is one of the strongest arguments for forex mean reversion as a portfolio strategy.

Practical Considerations

Forex mean reversion requires attention to execution details. Spreads widen during off-hours and news events, potentially eliminating thin edges. Rollover costs accumulate on positions held multiple days. Leverage magnifies both returns and drawdowns. We recommend using no more than 3:1 effective leverage even though brokers offer much more. The edge per trade is small enough that execution quality and cost management determine whether the strategy is profitable net of all frictions.

Conclusion

Forex mean reversion strategies work best on shorter timeframes (4-hour), during the Asian session, on range-bound pairs (EUR/USD, GBP/USD), using adaptive methods like z-score rather than fixed thresholds. The edge is real but thin, requiring high-quality execution and appropriate leverage. The greatest strength of forex mean reversion is the diversification benefit across multiple pairs, which produces a smooth equity curve with limited drawdown when run as a portfolio.