AlphaAssayVerdicts & the gradient
An AlphaAssay verdict is a diagnosis, not a yes/no oracle. Four outcomes exist:
pass, conditional, fail and insufficient_evidence —
and every one of them ships with the evidence that produced it.
The response, field by field
| field | meaning |
|---|---|
| verdict | pass · conditional · fail · insufficient_evidence. „Insufficient evidence" is an honest abstention — too little data to judge is an answer, not a failure of ours. |
| died_at | the first gate your signal failed (see the four gates). Absent on a pass. |
| failure_codes | machine-readable causes — your agent branches on these. Catalog: failure codes. |
| placebo_percentile | where your timing lands against 500 random twins with the same trading profile. 50 means „indistinguishable from chance". |
| survival_map | five live attacks — delay, cost stress, time jackknife, regime split, parameter wiggle — each marked survives / dead. |
| search_budget_left | how many honest tries your strategy family has left before multiple-testing eats the signal. Ends the endless tweaking loop. |
| signature | ed25519 signature over the canonical document — what makes the verdict independently verifiable. |
Why a fail is worth paying for
A no without a finding is an annoyance. A no with died_at and failure codes is a shortcut:
it tells you whether to fix the costs, stop tweaking the family, or abandon the direction entirely —
before real money finds out for you.
Determinism
Same input, same verdict, bit for bit, every time. That is what makes replay, caching and third-party verification possible. There is no model temperature anywhere in the battery.