DeflatedSharpeRatio<T>
Helpers & Utilities · AiDotNet.Finance.Evaluation
López de Prado’s Deflated Sharpe Ratio (DSR): the probability that a strategy’s true Sharpe ratio is positive, after deflating the observed Sharpe for (a) the number of strategy configurations tried (multiple-testing / selection bias) and (b) non-normality (skew and excess kurtosis) of returns.
For Beginners
If you try 1,000 random trading rules, one of them will look great purely by luck. The DSR corrects for that: it asks “given that I tested N strategies, how surprising is this Sharpe ratio, really?” and answers with a probability that the strategy’s edge is real (true Sharpe
0). It also accounts for the fact that real returns have fat tails and asymmetry, which make a raw Sharpe ratio less trustworthy. Higher DSR = more confidence the result is not a fluke.
How It Works
When many strategies are backtested, the best observed Sharpe is upward-biased simply by chance. The
DSR compares the observed Sharpe against the expected maximum Sharpe that N independent
trials of a truly skill-less process would produce, and expresses the result as a probability in
[0, 1] using the Sharpe-ratio estimator’s standard error (which itself depends on skew and kurtosis).
A DSR above ~0.95 is the usual bar for declaring a discovery genuine.
Methods
| Method | Summary |
|---|---|
T Compute(double observedSharpe, int nObservations, int nTrials, double skew, double kurtosis) |
Computes the Deflated Sharpe Ratio. |
T ExpectedMaxSharpe(int nTrials) |
Expected maximum Sharpe ratio (per-observation, i.e. |