Research methodology

An opportunity earns its place in the feed.

SignalRankAI combines market-data checks, strategy candidates, calibrated model decisions and risk admission. The absence of a trade is a valid result when evidence does not support execution.

REJECT WEAK EVIDENCE · PRESERVE STRONG RULES
01Discovery

Multiple markets, one source of truth.

Crypto, foreign exchange, indices, equities and commodities have different venue hours, instrument identities and data-quality requirements.

  • Identify canonical instruments before comparing provider symbols.
  • Validate freshness, coverage and available trading session.
  • Fail closed when the quote or venue is ambiguous.
02Decision

Strategy consensus before distribution.

Candidates pass defined scoring, strategy, regime, ML calibration and risk gates. A high-looking raw score alone does not certify a signal.

  • Identify strategy and timeframe alongside entry and invalidation levels.
  • Separate raw probability from calibrated probability.
  • Record the reasons rejected candidates did not qualify.
03Evidence

Delivery and trade lifecycle stay distinct.

Delivered signals can include proof of access, model quality context, market observations and an outcome ledger. These are not interchangeable with broker execution receipts.

  • Track receipt channel and confirmation timestamps.
  • Label paper fills as simulations, not live fills.
  • Confirm actual fills, fees and exits with the broker before treating an order as executed.
Disclosure. No historical backtest, ML probability or previous win rate guarantees a future outcome. Live decision and execution authority are governed by separate safety checks.
Continue reading: The risk system Open secure account