Mean reversion
Hypothetical EUR/USD H1 Bid closes have a mean μ = 1.1020 and sample standard deviation s = 0.0010. The latest completed Bid close of 1.1000 has z-score (1.1000 − 1.1020) / 0.0010 = −2. The available entry Ask is 1.1002. A target of 1.1020 Bid and SL of 1.0980 Bid imply 18 pips to the target and 22 to the stop before costs and slippage. Statistical deviation and trade parameters describe different quantities; −2 does not guarantee reversion.
What should revert, and to which reference?
Specify the quantity: price, return or a spread between instruments, together with the reference calculation. A moving average, fixed level and estimated pair relationship produce different hypotheses. Record the window, price feed, dispersion measure and update time. If an average follows a trend, a later touch need not represent price returning to its former level. An ordinary currency chart establishes neither stationarity nor a fixed equilibrium value.
Turn deviation into a complete rule
A z-score threshold or position relative to bands is a condition to assess, not an automatic instruction to buy. Define entry, waiting period, target, invalidation and treatment of further divergence. A momentum filter may form part of a variant, but its value needs testing. Adding to losses is not required by mean reversion; any scaling policy needs assessment alongside aggregate exposure and a no-reversion scenario.
Edge, risk and changing conditions
Repeated small gains can coexist with large losses, but the strategy family’s name establishes neither that distribution nor a high win rate. Calculate net results and include cases that did not revert within the specified period. Do not examine only successful rebounds. Changing conditions can weaken a historical relationship, while the estimated average is uncertain. A loss limit closes a position; it does not disprove the entire concept of mean reversion.
Further learning and sources
8.7 Trend, breakout and mean reversion: three families of entry logic; 5.8 Oscillators and momentum indicators—RSI, stochastic, MACD and ROC; 15.4 Backtesting and forward testing — methods, evidence and pitfalls.
