Track model rank performance over time using custom scoring
Custom Score = (Corr Weight × Corr) + (MMC Weight × MMC)
Default: 0.75×Corr + 2.25×MMC
Percentile = (totalModels − rank + 1) / totalModels × 100, per round (best ≈ 100).
Ranks each round on the trailing N-round average of the Corr/MMC score, like Numerai's MMC20 / Corr60 leaderboard columns. "Per round" uses each round's own score.