Loading
Comp Rating v1 prioritizes recent result strength, FNCS resume, reliability, sample size, and consistency. Earnings and PR are treated as optional context, not invented ranking truth.
Weights recent event ratings most heavily, so current form moves the rating more than older results.
Rewards results in higher-tier events — finals and LAN-weight tournaments count for more than lower-stakes lobbies.
Scales confidence by sample validity: source-backed rows and larger match counts hold more weight than thin or unverified samples.
Favors stable placement across matches; high variance lowers the rating even when peak finishes are strong.
Ratings are directional and should be read alongside live form, event context, and recent placements.
The tournament score is not an overall reputation score. It judges one event row, then profile pages and cards aggregate those rows later. Placement is the spine of the model, points and elims add context, and data confidence controls how much of the result survives.
Percentile curve. Top finishes climb fast; bottom-field results turn negative.
Points per match against mode and tier expectations.
Elims per match matter, but cannot rescue a terrible placement alone.
Victory rate bonus with a small no-win drag.
Rewards stable output across matches and penalizes shaky samples.
Tier and lobby-strength context lift stacked lobbies.
Formula shape
raw = placement * .38 + points * .22 + elims * .16 + wins * .10 + consistency * .06 + fieldStrength * .08 - penalty
+/-/G = clamp(raw * fieldSizeFactor * confidenceFactor, -100, 100)