The Deflated Sharpe Ratio: Correcting for Selection Bias, Backtest Overfitting, and Non-Normality
Read the paperopens papers.ssrn.com in a new tab
What they found
The deflated Sharpe ratio is a practical tool: given a strategy's observed Sharpe ratio, the length of the track record, the skewness and kurtosis of its returns, and the number of strategy variations tried, it computes the probability that the true Sharpe ratio is above zero. The 'deflation' comes from comparing the observed Sharpe against the expected maximum Sharpe that a search over that many trials would produce by chance, and from widening the confidence interval for fat tails and short samples. A strategy that looks great can have a deflated Sharpe probability well below 50%.
What you can use
- You can compute, with a formula, how likely it is that your backtest's Sharpe ratio is real given how many things you tried.
- Fat-tailed and negatively skewed returns (most option-selling and carry strategies) need a much higher observed Sharpe to be believable.
- The number of trials is an input; if you did not track it, the tool will tell you to be pessimistic.
Caveats
Requires an honest count of trials and treats them as independent. Assumes the return series is stationary. SSRN version linked.
Tags: backtesting, sharpe-ratio, overfitting, multiple-testing
Summaries are our own reading of the paper, not the authors' words. Educational only, not advice. Discuss it in Book Club.