Interactive research tool · canonical contract

Can this alpha be certified before it fades?

Enter its initial Sharpe, economic half-life and how many candidates you screened. CAC compares the total evidence the signal can ever generate with the evidence your certification standard requires.

Below the certification frontier
4.00 years available/ 7.24 required

3.24 additional years of half-life needed at SR 2.00

CR0.55
information5.77 / 10.45 nats

Alpha Survival Frontier

Above the red curve, lifetime information meets the declared evidence requirement.

Click anywhere · drag the black point · hover the curve

To reach the frontier, keep SR 2.00 and raise half-life to 7.24 years, or keep the current half-life and raise Sharpe to 2.69.

Bonferroni · adjusted α 0.00050
A Sharpe-2 signal selected from 100 candidates needs a 7.24-year half-life to reach the default certification frontier.
Canonical Gaussian contract: 5% family error budget, 90% power, Bonferroni correction.
Method

Three quantities. One frontier.

The browser exposes the closed-form canonical calculation. Richer unknown-scale and data-driven certificates remain in the AlphaValue Python package.

01
Lifetime information

Evidence the signal can generate before exponential economic decay exhausts its useful life.

Strength × lifetime

Higher initial Sharpe and longer half-life expand the available information budget.

02
Required information

Evidence demanded by reliability, power and the search multiplicity declared before deployment.

Reliability × power × search

Screening more candidates tightens the per-candidate error budget and moves the frontier upward.

03
Certification

Compare the economic information budget with the information required by the deployment contract.

CR ≥ 1 reaches the frontier

Below one, the declared lifetime is too short under the chosen canonical contract.

What CAC asks

A different question from significance.

A significance test asks whether observed performance clears a statistical bar. A track-record calculation asks how long a stable signal needs. CAC asks whether a decaying signal has enough lifetime information to clear the bar at all.

Observed significance

Did the observed signal clear the bar?

Ask whether the performance already observed is distinguishable from luck, possibly after accounting for selection.

Track-record length

How long would a stable signal need?

Ask how much observation time a signal of constant strength needs to reach the chosen evidence standard.

Certified Alpha Capacity

Is certification possible before decay wins?

Ask whether the entire remaining economic life of a decaying signal contains enough information to reach the evidence requirement at all.

Canonical scope. The interactive threshold uses the paper's Gaussian, known-volatility, exponential-decay, fixed-sample contract. Reaching the frontier means certification is feasible under that contract; it is not a profitability claim.

Theory and illustration

Mathematical guarantees, numerical constructions and descriptive evidence answer different questions.

01
Learning before decay

A reliability frontier in a specified statistical experiment.

The canonical Gaussian model puts statistical learning and opportunity decay on a common information clock. Its feasibility frontier determines whether the declared reliability contract can be met.

The guarantees require the likelihood and observation assumptions in the paper. Cash value also depends on the payoff map and its uniform regularity conditions.

02
Funding persistence

A descriptive empirical illustration.

The funding panel examines how a lifetime score relates to persistence in future signed funding, including information beyond contemporaneous funding magnitude.

This association does not establish executable hedged-trade returns. No confirmatory trading-profit validation is claimed.

03
Learning near the frontier

Residual information and cash payoffs have distinct asymptotic scales.

Near-frontier figure comparing the inverse-logarithmic asymptotic reference, feasible two-look values and a fixed-time benchmark, with a separate comparison of information and cash exponents.
The asymptotic reference shows the sharp limiting law. Cash exponents use the uniform reward-family assumptions in the paper. Open figure at full size ↗
Research artifact

Read it. Inspect it. Reproduce it.

The calculator is one surface of a theorem-driven release with source, tests and a reproducibility record.