On paper, a Sharpe ratio of 2 looks excellent. Most professional funds would be thrilled with a sustained Sharpe near 1. So a backtest showing 2 feels like you've found something rare. Maybe you have. But whether that number is real skill or a statistical mirage hinges on one question almost nobody asks.
The Sharpe ratio is return per unit of risk — roughly, how much reward you got for how bumpy the ride was. As a rough field guide on annualized numbers:
The catch: those benchmarks assume the number came from one honest test. Almost no backtest does.
Test one strategy and get a Sharpe of 2 — impressive. Test a hundred variations and keep the best one that hit 2 — that's not skill, that's the loudest coin flip. With enough attempts, a high Sharpe appears by pure chance, and you fall in love with the survivor.
The deflated Sharpe ratio corrects your number for how many strategies you tried. The more variations you searched, the higher the bar your Sharpe must clear to count as real. A Sharpe of 2 from 1 test can be strong; the same 2 from 200 tests can be statistically meaningless. Same number, opposite verdict.
A paper Sharpe of 2 can fall below 1 — or below zero — once realistic trading costs are subtracted. If your backtest ignores costs, its Sharpe isn't measuring anything tradeable.
Mean Sharpe flatters you. Split the history into folds: a strategy can average 2 while one fold sits at −1. That worst fold is the regime where it actually blows up — and the average quietly buries it.
Ordinary cross-validation leaks in trading, because nearby trades overlap in time. Without purged cross-validation, your out-of-sample Sharpe is secretly part in-sample — inflated.
It's good if it survives all four questions: it holds after costs, across every purged fold (not just on average), and after deflating for the number of strategies you tried. If it survives that, a 2 is genuinely worth trading. If it doesn't, it was a beautiful number attached to nothing.
You don't need to reveal your logic, code, or parameters — only the per-period returns (e.g. 0.012, -0.004, 0.008 …) and how many variants you tried. From those alone you can compute the deflated Sharpe, run purged cross-validation, subtract costs, and check the tail — and get a blunt verdict: real edge, self-deception, or dead.
That's exactly what isitalpha does. You paste the numbers, we do the statistics, and your strategy stays yours.
Free verdict in 60 seconds. You share only the numbers — never your logic, code, or account.
Get an honest verdict →We ran 5,763 strategies through an honest gauntlet; most died exactly here — a high Sharpe that couldn't survive deflation and costs. Better to find out before you risk capital, not after.