Manual traders chase price. Algorithmic traders chase relationships. That distinction sounds subtle until you realise that price alone is noise — but the spread between two correlated instruments has structure, mean-reversion tendencies, and, critically, measurable statistical properties. That gap between manual intuition and quantifiable edge is exactly where statistical arbitrage lives.

Stat arb isn't a single strategy — it's a framework. At its core, the approach identifies pairs or baskets of instruments whose prices historically move together, then trades the divergence when that relationship temporarily breaks down. The algorithm doesn't predict direction; it bets on reversion to a modelled equilibrium. That's a fundamentally different risk profile to directional trading.

CONCEPTStat arb profits from mean reversion in price spreads — not from predicting which way the market moves.
WARNINGCorrelations that held for a decade can collapse in a week — regime shifts destroy stat arb models silently and fast.
KEY IDEASignal quality in stat arb depends on cointegration, not correlation — these are related but critically different concepts.

Here's where most newcomers stumble: correlation tells you two assets move together; cointegration tells you their spread is stationary over time. You can have two highly correlated assets whose spread drifts permanently — useless for arb. A cointegrated pair has a spread that reverts to a long-run mean, and that's the tradeable signal. Getting this distinction wrong turns a clever strategy into slow, expensive capital erosion.

Spread Mean-Reversion: Stat Arb Signal ZonesMean+2σ Sell-2σ BuyTime →Spread

Backtesting a stat arb strategy is where optimism goes to die — in a good way. Walk-forward testing, out-of-sample validation, and transaction cost modelling are non-negotiable. Overfitting a cointegration model to historical data is embarrassingly easy; the spread looks pristine in backtest, then immediately forgets how to mean-revert in live markets. For foundational reading, Investopedia's statistical arbitrage overview is a solid starting point, and Wikipedia's treatment of statistical arbitrage covers the academic lineage well. The underlying mathematics leans heavily on cointegration theory, which is worth understanding before writing a single line of strategy code.

Execution quality matters as much as signal quality — a mean-reverting spread erodes fast once transaction costs and slippage enter the picture. Build the model right, stress-test it hard, and assume the market will find your weakest assumption first.

This content is for educational purposes only and does not constitute financial product advice. Past performance is not indicative of future results. Profit Logic Ltd (ACN 688 669 936) accepts no responsibility for errors or omissions in this content or anywhere on this website. Always seek advice from a licensed financial adviser before making investment decisions.