Open research instrumentContract v1.0.0

How much Sharpe survives the search?

A strong backtest is less surprising after a long search. Put the observed Sharpe, return shape and complete selection process on the same surface. The calculator reproduces ALPHAC's per-period PSR and DSR formulas in your browser.

Open the pressure chamber Inspect the Python source

Illustrative calculation

The current-union preset supplies only trial count and dispersion. The observed Sharpe, sample length, skew and kurtosis are illustrative until you replace them. No output on this page is an ALPHAC performance claim or an admission verdict.

Selection pressure chamber

One result. The whole search behind it.

Observed sampleIllustrative until replaced Observed Sharpe, annualized Converted to per-period Sharpe before inference. Return observations Use the count after dropping missing returns. Periods per year 365 for daily crypto; 252 for trading days. Sample skew Negative skew can widen estimator uncertainty. Non-excess kurtosis Gaussian returns equal 3, not 0.

Selection processPreset from the public union Effective independent trials Do not shrink N after seeing outcomes. Cross-trial Sharpe SD, annualized Derived from the identity-aligned per-period V[SR].

Published union context 229 identities befea0cc80a42494...

Reducing the trial count or dispersion after observing outcomes flatters DSR. The selection union must be defined independently of the selected result.

CalculatingPending

Deflated Sharpe ratio ... Against the expected best result under the supplied search.

Observed...

Best by luck...

...Annualized Sharpe...

PSR against zero: ...
Best-by-luck benchmark: ...
Search haircut: ...
Non-normality term: ...

Read the result

PSR asks about the sample. DSR asks about the search.

PSR

Was the observed Sharpe above a benchmark?

Probabilistic Sharpe Ratio adjusts for sample length, skew and non-excess kurtosis. The zero-benchmark result does not know how many alternatives were tried.

SR*

What would the best null trial look like?

The expected maximum rises with the number of effectively independent trials and their cross-trial Sharpe dispersion.

DSR

Did the result clear the search-adjusted benchmark?

DSR is PSR evaluated against SR*. It is one diagnostic inside a larger research contract, never proof that a strategy is deployable.

Implemented formula Phi(((SR - SR*) * sqrt(T - 1)) / sqrt(1 - skew*SR + ((kurtosis - 1)/4)*SR^2)) sqrt(V[SR]) * ((1-gamma)*PPF(1-1/N) + gamma*PPF(1-1/(N*e)))

sr_per_period = sr_annualized / sqrt(periods_per_year). variance_per_period = (sd_annualized / sqrt(periods_per_year))^2.

Current ALPHAC policy

A probability is not a promotion decision.

Under canli.alphac-sleeve-admission-contract.v7, per-sleeve DSR is mandatory to measure and publish, but it is not a universal sleeve gate. Incremental admission is not decided by DSR alone.

The 0.95 reference applies to a full-union portfolio-maturity claim. Clearing it inside this calculator does not clear robustness, execution, stress, capacity, bootstrap-improvement or trial-accounting requirements.

Read the complete policy context

Reproduce and inspect

The interface is downstream of the evidence.

Formula contract241f33d627cecb87...Golden vectors and policy boundary Trial unionbefea0cc80a42494...Current N and identity-aligned V[SR] Python implementationac2f18440e21b824...Production arithmetic in ALPHAC Primary paperBailey and López de PradoJournal of Portfolio Management 40(5), 94-107 (2014)

This contract reproduces ALPHAC's PSR and DSR arithmetic for supplied inputs. It does not validate a strategy, prove profitability, estimate current portfolio performance, or replace the complete admission contract.