AlphaAssay — Trading Signal Validator

A thousand signals promise an edge. Most are noise. Is yours?

AlphaAssay is an independent assay office for trading signals. Send a strategy, a backtest or a returns series, and it goes on a statistical trial — out of sample, against chance, against overfitting — and comes back with a deterministic pass or fail verdict, with the reasons attached. Every verdict is replayable, bit for bit. Proof, not promises.

This Actor runs that trial from inside Apify: pick one of the built-in test cases or submit your own signal, and get the full verdict as structured output your agent can branch on.

Why this exists

The uncomfortable truth of quant research: most profitable-looking strategies are statistical illusions. When 97 peer-reviewed return anomalies were re-tested, more than half of the performance evaporated after publication. When 2.1 million strategies were generated systematically and corrected for multiple testing, almost none survived. Retail backtests get far less scrutiny than either — and then real money follows them.

A no without a finding is an annoyance. A no with a finding is a shortcut: it saves the most expensive thing you own — weeks of searching in a dead direction, and the drawdown that ends the search for you.

What the trial does

Every signal runs the same four-gate battery, in order:

1. Net edge — does anything remain after realistic costs, spread and one bar of execution delay? Most edges end here.
2. Family deflation — every variant you tried is counted against you. Trying forty parameter combinations and keeping the best one manufactures a great backtest from pure noise; the deflated Sharpe ratio prices exactly that.
3. Placebo trial — your signal races 500 random twins with the same trading profile. If random timing does as well, your timing was never the edge.
4. Robustness — the survivor is attacked: delayed fills, stressed costs, deleted history, regime splits, wiggled parameters. A real edge is inconvenient to kill.

The verdict names the first gate that killed the signal — so you know what to fix, or when to stop.

Input

Field	What it does
specimen	One of four built-in known-answer test cases (default: golden_lookahead, a signal with planted look-ahead bias the battery must catch)
customPayload	Your own signal instead: a returns series, equity curve or trade list as JSON

The built-in specimens are the fastest way to audit the validator itself: each has a known correct verdict. A validator you cannot test is just another promise — so test this one first.

Output

The complete verdict document, pushed to the dataset:

{
  "schema": "gauntlet.v1",
  "verdict": "fail",
  "died_at": "net_edge",
  "failure_codes": ["no_net_edge"],
  "spec_hash": "sha256:421f1b1f4675…",
  "family_id": "fam_004b8867ffc7a556",
  "stages": [ … ]
}

• verdict — pass, conditional, fail, or an honest insufficient_evidence when the data cannot support a judgment either way.
• died_at — the first failed gate: net_edge, family_deflation, placebo or capacity.
• failure_codes — machine-readable causes. An agent that branches on these converges; one that retries blindly burns its budget on a dead idea.
• stages — the full evidence trail, gate by gate.

Pricing

The four golden specimens are free — they are known-answer tests that prove the validator is correct before you trust it with your own data. Validating your own signal (the Custom input) is charged per event at the price shown on this page; pass and fail cost the same, because you are buying the trial, not a flattering outcome. Platform pricing for direct API access: alphaassay.com/pricing.

Don't trust us — check us

• Verdicts on the paid tier are Ed25519-signed; anyone can verify a certificate at alphaassay.com/verify or completely offline.
• Our own track record is public and updated daily — we grade ourselves before we grade you: GET https://api.alphaassay.com/v1/public/calibration.
• Strategies are never stored. Only a one-way fingerprint and the cause of death.

Links

• Website: alphaassay.com
• Documentation: alphaassay.com/docs
• MCP server and known-answer examples: github.com/alphaassay

---

AlphaAssay performs methodology audits. It is not investment advice, has no order path and takes no custody. A pass means a signal survived a statistical trial that most signals fail — it is never a promise of returns.