Kalman Box-Breakout Stat-Arb

Family: mean_reversion · Regime: ranging · Complexity: high · Asset classes: Crypto (ETH/BTC), Major FX Pairs (AUD/NZD, EUR/GBP) · Timeframes: 15m, 1h

Thesis

Statistical arbitrage via Kalman-filtered spreads generates alpha in non-trending regimes; however, it often fails during 'drifting' trends or low-volatility traps. By filtering for low ADX (non-trending) but requiring a Boxline breakout (volatility expansion), we isolate high-probability mean-reversion entries where price is actually moving back toward its statistical equilibrium rather than oscillating in a dead zone.

Components

• Average Directional Index (regime) — Used to ensure the market is not in a strong directional trend, as statistical arbitrage/mean reversion underperforms in trending regimes.
• QQE (Qualitative Quantitative Estimation) (direction) — Confirms momentum in the direction of the mean-reversion before entry to avoid catching a 'falling knife'.
• AI Pair Trading System v1.0 (entry) — The core entry signal; uses Kalman-filtered Z-scores and ML weighting to identify statistical mispricings between paired assets.
• Restu Kaioh Scalping Indicator (exit) — Uses stochastic exhaustion to exit trades when the mean reversion has likely reached local momentum maturity.
• Pivot Points Reversal Levels (risk) — Provides dynamic support/resistance levels based on local two-bar reversals to set tight, market-structure-based stop losses.
• Boxline (Consolidation Zones) (volatility_filter) — Acts as a volatility expansion filter; entries only occur after a box breakout to ensure sufficient price movement for the mean-reversion signal.

Known failure conditions

• Pair correlation coefficient drops below 0.7 over the lookback period.
• ADX stays above 35 for sustained periods, indicating a regime shift to trend-following.
• Frequent SL triggers at Pivot Levels without reaching the mean, suggesting the Kalman mean is lagging price action significantly.

Explore the full interactive blueprint with parameter ranges and evidence on WOBR StrategyVerse, or generate this strategy as an MT4/MT5 Expert Advisor with QuantMogul AI Engine (free download).