PCA Statistical Arbitrage — Research Tearsheet

Eigenportfolio Mean-Reversion with Continuous Sizing | BN 35c, 2022-07 → 2025-11 | Generated 2026-03-20

Key Metrics — BN 35c (3.4yr)

Sharpe
1.604
Sortino
2.361
CAGR
16.7%
MaxDD
-7.3%
Calmar
2.27
Win Rate
55.0%
Profit Factor
1.35
BTC Corr
-0.19
BTC Beta
-0.03
MC P(>0)
99.9%
MC P(>1.0)
90.5%

Cross-Dataset — HL 31c (5mo)

HL Sharpe
1.893
HL Return
5.4%
HL MaxDD
-2.3%

Strategy Summary

PCA extracts 7 common factors from 35 crypto perpetuals. Coins whose residual s-score deviates beyond threshold are mean-reverted via continuous Carver/AHL-style sizing. BTC trend filter + inverse vol scaling adapt to market regime.

Key Properties: Market-neutral, fee-immune (continuous sizing + 25% deadband), uncorrelated with trend-following.

Production Config: 7 PCs, 504h lookback, 72h s-score window, decay 0.995, vol_target 0.25, top_k 3, rebal 4h, entry_z 1.8