Metodoloji · sahnelenmiş
Metodoloji v2
v2 resolves supported crypto, stock, metal, and forex signals from historical provider candles at the exact 24-hour and 7-day target horizons. Provider misses stay pending and are not counted as wins or losses.
What changes vs v1
- Resolution coverage: crypto uses Binance hourly klines. Stocks, forex, and metal futures use Yahoo hourly chart candles. The first candle closing at or after the target timestamp is selected.
- Venue coverage: only exchanges whose market data Signodex is licensed to redistribute are resolved. Historical signals on venues that have since been removed stay in the ledger but no longer resolve, and are excluded from the denominator rather than scored against an unlicensed price.
- Thresholds: unchanged. Bullish accurate at ≥+0.5%, bearish at ≤-0.5%, neutral at |Δ| < 2.0%. Wilson 95% CI canonical, point estimate is the midpoint.
- Aggregation: only successfully resolved horizons enter the denominator. Source and observed timestamp are stored with every resolution.
Resolution timing
A scheduled resolver runs after a horizon becomes eligible. The job's execution time does not determine the evaluation price: it queries history and stores the provider candle timestamp used for scoring.
Corrective backfill
Values produced by the previous retrieval-time quote resolver are invalidated before v2 is enabled. Eligible rows are recomputed from historical candles. Until a row is recomputed it remains pending and cannot affect the published accuracy rate.
Audit hash
The thresholds in §1 and the resolution-source list in §1 jointly produce v2's identity. A change to either requires
methodology_version = 3 and a new page. The corrective provenance migration is 043_historical_outcome_provenance.sql. Git history is the audit trail.Past performance does not predict future performance. See Disclaimer. For live numbers see track record.