EN
EnglishEspañolPortuguês
My accountAudit
Back to the home page/Español/Português
Free calculator

Deflated Sharpe ratio calculator

Explore the deflated Sharpe ratio using declared Sharpe, years of history and research attempts. Assumptions and limits are explicit.

Your numbers

The one your platform shows, annualised. For example 1.8.

From the backtest's first to its last day. For example 3 or 0.5.

Every parameter combination you ran, every version you dropped and every optimizer pass. If you don't know, give your best estimate; it is usually more than it seems.

Use your series frequency. This is a declaration, not a measurement.

With your file, the figures are measured

The calculator uses your declared frequency and assumes normal returns. Your file has real skew, tails, costs and data gaps, and the report measures them, along with the number of MT5 optimizer passes when you upload it. Free and without an account, you see your file's A to D class and red flags. The full report, with every figure and the PDF, costs USD 29. If it is no use to you, we refund your money: ask within 7 days of paying.

Audit my file See a sample report

Read next: Deflated Sharpe ratio for track records · Create a shareable figures card

Why the search matters

  • Try 100 configurations with no real edge and the best of them almost always shows a high Sharpe. That is the one you pick and the one you publish.
  • The deflated Sharpe compares your Sharpe with the one the best of N random tries would show. The more configurations you try, the higher it must be to count.
  • More history helps: the spread of luck shrinks with the years, so a longer backtest needs less Sharpe to beat the same search.

Assumptions and limits

  • Monthly · 12 periods/year (Declared). Returns are assumed to have no skew and normal tails. Fat tails and negative skew make the real result stricter.
  • Configurations are treated as independent tries with the same spread as your Sharpe.
  • Harvey and Liu haircut with the Bonferroni correction; expected luck from Bailey and López de Prado; minimum length from Bailey, Borwein, López de Prado and Zhu.
  • We keep your figures in no database or file; since they travel in the link, the server's access log may hold the address requested. It is the same formula as the report's luck section; the report uses your file's actual frequency.
  • It describes the past you declare; it says nothing about the future.