Free tool
The detection floor on your backtest
A short track record often cannot support the edge printed on it, and here is the math. Enter how long your sample is and how many variants you tried, and see the smallest edge a rigorous gate could even certify at that sample size. Everything runs in your browser; nothing is sent anywhere.
If your browser has JavaScript turned off, this calculator cannot run. The same math is available in our command line tool and on every certificate we issue.
Why this number matters
Any backtest can print a high Sharpe. The question is whether that number could survive its own estimation error. Over a short sample, the noise in a measured Sharpe is large, so only a very big edge could be told apart from luck. This floor is the smallest incremental Sharpe whose probability of being a real edge, after accounting for selection luck, reaches 95%. It is set by your sample length, not by how good your strategy looks.
Two honest consequences fall straight out of the math. A 90 day daily backtest can only certify a very large edge, so most short-sample strategies cannot be proven at all, whatever their in-sample Sharpe. And trial count alone does not lower or raise the floor: deflation needs to know how varied your variants were, not just how many there were, which is why the spread field starts at zero and stays inert until you fill it in.
This is the exact closed form the paid audit runs. The browser code is a faithful port of the Python gate, and a test in our repository fails if the two ever disagree. See the full overfitting audit for a certificate on your own returns, or the kill log for the same gate applied to our own research.