Evidence, calibration & model readiness
Both competitions are now in data collection and model-validation probation. No signals will be promoted until out-of-sample evidence is strong enough. Open roadmap →
Historical results and clean V4 validation are shown separately. An active prediction is never counted as a win, and small samples are not promoted as proof.
Observed segment only. Promotion still requires a larger settled sample and selection-level evidence.
Bands with very small n are displayed but are not treated as calibrated proof.
Live trading is disabled. The system can withhold all preferred signals when evidence is insufficient.