Mean Reversion

Prices oscillate around value in RANGE regimes; a 2σ stretch (ZSCORE < -2) in such a regime is a rubber band. The fatal error is fading extremes in TRENDS — trends are the regime where extremes extend. Hence this platform's stat-arb entry demands both: stretch AND a mean-reverting regime (Hurst < 0.4).

Hurst < 0.4 / ADX < 20 environments. The regime filter isn't decoration — it IS the strategy.

STAT_ARB_ENTRY_LONG: z-score ≤ −2 while HURST_EXPONENT(100) < 0.4. Optional candle confirmation (hammer) at the extreme.

At the mean: ZSCORE > 0 (the preset's exit). Mean reversion has a natural target — overstaying past it converts the strategy into hoping.

Hard stop ~7% or 2.5σ — because occasionally the 'extreme' is actually a regime change beginning, and that tail must be amputated.

Risk ≤1%; because win rate is high and losses rare-but-larger, sizing discipline matters MORE here, not less.

The P&L shape is inverse to trend following: many small wins, occasional sharp loss. Never remove the stop 'because it always comes back.' It doesn't, once.

Short-horizon reversal is well documented in equities; the preset pairs the entry with its regime filter and honest stop — backtest it against a trender and a chopper and compare. Every strategy here maps to this platform's Backtest panel — the matching Strategy Preset fills entry/exit/stops in one click, and every result is benchmarked against buy-and-hold, which most simple strategies LOSE to on strong trenders. Seeing that honestly is the lesson.