audit2 — candidate & methodology library

The positive side of /audit3's kill tribunal: what is worth forward-testing, and the measurement kit to judge it. Nothing here is money-ready.
population: —

Doctrine

How this surface thinks about edge
“Public daily OHLCV is arbitraged by thousands of participants. Remaining public edges are (i) risk premia and (ii) microstructure/flow. Everything else is noise dressed as signal.”
“Volatility scaling is free alpha. Simply scaling positions by inverse realized vol improves Sharpe by 0.3+ for any momentum strategy.”
“Edge, if any exists, should be expressed with symmetric, vol-scaled, time-based exits and sized by vol targeting — the expression of every trade changes the statistics more than the signal does.”
“The binding constraint is hypothesis quality and trial budget, not techniques.”

Verdicts (from audit3_ledger.json)

Asset classVerdictnNet expectancy/tradeNet 95% CI2× cost OKDSRrequired_n
—
2× cost OK = the net-expectancy CI still excludes zero at twice the class slippage (a required gate for PROVEN_T2 / CANDIDATE). required_n = trades needed to confirm the observed effect at 80% power.

Candidate & observe pods

Pre-registered before forward data. pre_registered may confer confirmatory status; observe cannot promote until promoted to pre_registered.

Forward evidence clock

A pod can only promote on forward rows dated after its registration. Rows are aggregated per hypothesis from reports/forward_observations.jsonl into audit2_forward.json; the owning pod and the gate reasons are joined in from audit2_pods.json and the ledger's forward_gate.
HypothesisOwner podForward rowsBenchmark leg (coverage)Intrabar-resolvedStatus
—
—

Trial budget & benchmarks

—
ClassBenchmark legGate
Cost gate: average net PnL must be at least 3× the measured slippage. ETF slippage measured ≈234% of per-trade PnL (monthly cadence only); CRYPTO ≈37%.

Measures library

MeasureWhat it is

Traps / graveyard

IdeaWhy it is not a candidate

Retired with reason

ItemReason

Provenance & freshness

—