Mean reversion · Intermediate · Daily monitoring — positions can be held days to weeks
Pairs Trading
How it works, and when it doesn't.
Correlation is not a static number. The trade profits when a temporary divergence in correlated assets reverts — and history shows it almost always does.
Spread z=-1.42 within ±1.5σ. β=1.015. Spread=-0.0016. No mean reversion trade.
Why it works
Pairs Trading exploits the mean-reverting spread between two correlated assets. When SPY and QQQ diverge beyond 2 standard deviations of their historical spread, it implies one is temporarily mispriced — buy the cheap one, short the expensive one.
The signal logic
- Calculate rolling 60-day z-score of SPY/QQQ spread
- Long the laggard, short the leader when z-score > 2.0
- Exit when spread returns to mean (z-score < 0.5)
When it fails
When the correlation breaks permanently due to a fundamental change. Tech underperforming for years (2000-2002) would cause endless losses on a QQQ-SPY pair.
- Best market
- Range-bound markets, stable correlation regimes
- Worst market
- Sector rotation periods, regime changes that break correlations
Backtest
| Sharpe ratio | 0.28 |
| Annual return | 0.6% |
| Worst drawdown | -3.0% |
| Trades in the backtest | 106 |
Backtest on daily bars, positions taken the bar after the signal (no look-ahead).
Pairs Trading, live.
Re-scored after every close.
The platform runs this and 32 other strategies against the live tape each day — with the evidence behind every number.