Robustness
The Deflated Sharpe Ratio (DSR): the Sharpe that doesn't lie
7 min read · by the GetBacktest team
The Sharpe ratio measures return per unit of risk — but it has a major flaw in backtesting: if you test a hundred strategies, the best will have a fine Sharpe by pure chance. The Deflated Sharpe Ratio (DSR) corrects exactly that bias. It's one of the few numbers that tells a real edge apart from selected luck.
A reminder: the Sharpe ratio
The Sharpe relates average excess return to its volatility: the higher it is, the better the risk-adjusted return. It's the most widespread performance measure.
Its limits: it assumes roughly normal returns (yet trading strategies often have fat tails) and, above all, it says nothing about how many trials it took to obtain it.
The multiple-testing problem
If you try a single strategy and it has a good Sharpe, that's meaningful. If you try a hundred and keep the best, its Sharpe is inflated by selection: you picked a lottery winner, not an edge.
That's selection bias (or data snooping): the more attempts and parameters you multiply, the easier it becomes to surface a flattering curve down to chance.
What the DSR corrects
The deflated Sharpe adjusts the observed Sharpe for the number of strategies tested, the length of history, and the shape (skew, kurtosis) of returns. It answers: “is this Sharpe credible given everything I tried?”
The result reads like a probability that the edge is real rather than down to chance. A high DSR is far more reassuring than a high raw Sharpe obtained after a thousand trials.
How to use it
Treat the raw Sharpe as a first impression and the deflated Sharpe as the verdict. A good Sharpe that collapses once deflated signals a strategy that's probably overfitted or over-selected.
On GetBacktest, the deflated Sharpe is part of the robustness verdict, alongside walk-forward and Monte-Carlo. The goal is the same: believe an edge only when it survives statistical scrutiny.
Don't believe it — prove it
Backtest this concept on real data, tick by tick, and get a robustness verdict. 7 days of Pro free, no card.
Start for freeFrequently asked questions
What is a good Sharpe ratio?
In trading, an annualized Sharpe above 1 is decent, above 2 very good — but a high raw Sharpe is worthless if selected among hundreds of trials. Always look at its deflated version.
Why “deflate” the Sharpe?
Because testing many variants mechanically inflates the best Sharpe by chance. The DSR removes that inflation to estimate whether the edge is real.
Does the DSR replace walk-forward?
No: the DSR corrects selection bias, walk-forward tests out-of-sample performance. Both, plus Monte-Carlo, form a complete robustness verdict.
Read next