Pair Trading Cointegration Tester

Z-score-based mean-reversion entry/exit on a candidate pair, with parameterized lookback, entry threshold, exit threshold, and stop-loss. Outputs include realized Sharpe, max drawdown, trade count, and win rate. The methodology page documents every parameter.

It is — the tester requires the pair to pass an Engle-Granger test before running. If the pair fails cointegration, the test refuses to run rather than producing misleading numbers. Use the Cointegration Half-Life Solver to identify cointegrated pairs first.

The pair tester operates on bar data (1m/5m/daily) and assumes mid-quote fills. The order-book replayer is tick-level with realistic fill simulation. Use the pair tester for strategy ideation; use the order-book replayer to validate execution feasibility before sizing.

Higher threshold = fewer trades but higher conviction. Common starting points: enter at z=2 (about 5% tail), exit at z=0 (mean), stop at z=4. The tool reports trade count and Sharpe per threshold, so you can sweep and see the surface.

Three reasons covered in the methodology page references: (1) cointegration breaks during regime changes, leaving you with a divergent pair, (2) execution costs eat the alpha on tight pairs, (3) the strategy is well-known and crowded, so edges have compressed since the early 2000s. The tester is for studying the strategy, not for guaranteeing profit.