Noise Injection Robustness Tests

A backtest is really just one draw from an enormous space of possible price paths and possible parameter choices. If a strategy's Sharpe ratio collapses the moment you nudge the input data by an amount smaller than typical bid-ask noise, the strategy was never really finding a pattern, it was fitting the exact historical sequence. Noise injection tests this directly: add small random jitter to prices, signals, or fill assumptions, rerun the backtest many times, and look at how much the performance numbers move.

How it's done

Pick a noise source, commonly a small Gaussian shock added to each day's return, or a random one-tick shift to each fill price, and re-run the full backtest dozens or hundreds of times with a fresh noise draw each time. A robust strategy shows a tight cluster of Sharpe ratios and drawdowns around the original result. A fragile, overfit strategy shows results that swing wildly, or even flip from profitable to unprofitable, under perturbations far smaller than real-world execution slippage.

Worked example

A strategy backtests at a Sharpe ratio of 1.8. Adding noise equal to one basis point per trade, smaller than the typical bid-ask spread, and rerunning 200 times produces Sharpe ratios ranging from 0.2 to 2.4, with many runs near zero. That spread is a red flag: the strategy's edge depends on hitting exact historical prices, not a pattern that would survive contact with real execution costs.