Autoencoder reconstruction error stat arb
RL/MLNeutral
Train autoencoder on returns; trade stocks with high reconstruction error (anomalies) to mean revert
Source: ML Quant
Box-Tiao canonical analysis
Multi-AssetNeutral
Find linear combos with most predictability; canonical correlation method
Source: Box-Tiao 1977
ETF-vs-basket spread
Multi-AssetNeutral
Spread ETF vs synthetic basket of constituents; trade NAV deviations intraday
Source: Avellaneda
Factor-residual stat arb
Multi-AssetNeutral
Regress out Fama-French/sector factors; trade idiosyncratic residual mean reversion
Source: Academic Lit
Graph neural network pair discovery
RL/MLNeutral
GNN learns pair relationships from historical data better than static cointegration
Source: Recent ML
Hierarchical Risk Parity allocation across pairs
Multi-AssetNeutral
Allocate capital across pairs using Lopez de Prado HRP; better than equal weighting
Source: Lopez de Prado
LSTM spread prediction
RL/MLNeutral
Train LSTM on spread history to forecast 1-5 day reversion direction; use as filter
Source: ML Quant
Principal components mean reversion
Multi-AssetNeutral
Trade residuals from regressing stocks on top-k PCs; Avellaneda-Lee 2010 framework
Source: Avellaneda 2010
Reinforcement learning stat arb (PPO)
RL/MLNeutral
Train PPO agent on spread state to learn entry/exit thresholds vs static z-bands
Source: ArXiv 2403
Sector-ETF rotation stat arb
Multi-AssetNeutral
Cross-sectional residual mean reversion within sector ETF universe
Source: Quant
Sparse PCA stat arb
Multi-AssetNeutral
Sparse PCs interpret as sector factors; residuals drive trades
Source: ML Quant
Sparse mean-reverting portfolio (SMRP)
Multi-AssetNeutral
Optimize weights to construct sparsest mean-reverting linear combo via SDP/iterative LP
Source: Zhang 2020
Stat arb on sector-neutral baskets
Multi-AssetNeutral
Build long-short baskets neutral to sector dollar exposure; capture residual alpha
Source: L/S Equity
Stat arb with Granger-causal lead-lag
Multi-AssetNeutral
Trade lagger when leader moves; identify via Granger causality tests
Source: Quant Lit
XGBoost meta-label on cointegration signals
RL/MLNeutral
Use ML to predict win/loss probability for each cointegration entry; trade only high-conf
Source: Lopez de Prado
An encyclopedia of publicly documented strategies for education. YCAI does not recommend, endorse, or trade any of these. Not investment advice.