Mizuki

Scores are normalized within the fighter's current division. Small samples are shrunk toward average instead of being allowed to dominate the leaderboard immediately.

Ratings refresh after a complete verified fight-stat card is available. Earlier snapshots stay preserved.

These are the most direct opponent adjustments behind the current score — performance against each opponent compared with what that opponent normally produces or allows in the same division.

Fight-level records are used for model context and raw-stat verification.