Validation method

Deflated Sharpe ratio

DSR · Bailey and López de Prado, 2014

The deflated Sharpe ratio asks how much of an observed Sharpe remains after many trials and non-normal returns.

What the ratio is for

A backtest Sharpe is easy to inflate. One source is the search itself: many rules are tried, and the one that is shown is the one that looked best. Another source is the shape of returns. Skew and fat tails change how surprising a Sharpe is. Bailey and López de Prado (2014) put both into one reading, the deflated Sharpe ratio. A Sharpe that looks rare on its own can be ordinary once the search that produced it is counted.

How Alphavo reports it

After exits have been compared, Alphavo writes the result as one of three words: high, medium, or low. High is the stronger label. Medium sits between. Low is the weaker label. The published result is the word. It is not a reject line. It is not a p-value. It is not a 0.95 cutoff.

What the label does not do

The label does not assign 0 stars, 1 star, 2 stars, or 3 stars. Those stars appear when a strategy is shown. A low label is not itself a decision to drop the candidate. A Candidate Strategy is not Alpha.

Common questions

What is the deflated Sharpe ratio?

The deflated Sharpe ratio, from Bailey and López de Prado (2014), asks whether an observed Sharpe still looks high after the number of trials and the skew and fat tails of returns. A Sharpe that looks rare on its own can be ordinary once the search that produced it is counted.

Does Alphavo reject a strategy when the deflated Sharpe ratio is low?

No. Alphavo writes the result as high, medium, or low. The word is not a reject line, not a p-value, and not a 0.95 cutoff. It does not set the star.

Is a high deflated Sharpe ratio a probability of future profit?

No. The label records how the observed Sharpe sits once the trials are taken into account. It is not a forecast that the candidate will keep making money.