NHL RAPM Ratings · 2016-2017
Methodology →Offense, defense, and total 5-on-5 shot impact per 60 minutes, relative to league average. A rating of 0 is exactly average; higher is better for all three. Sort any column; search by player or team.
2016-2017Games1,230Players881Eligible666Ridge λ31.4Coverage dispersion0.188
Coverage dispersion 0.188 — the fifth-most-complete of 9seasons. Higher dispersion means the public play-by-play undercounts some events, which can shift a season's overall rating level. Compare rankings within a season; compare levels across seasons with caution. Methods →
How to read these ratings
- Units:Shots per 60 minutes of 5-on-5 ice time, relative to the league average that season.
- Sign:Positive = above league average, negative = below, 0 = exactly average. Higher is better for Off, Def, and Total (Def is oriented so shot suppression reads positive).
- Scale:Around 0 is league-average. The top 10% of eligible skaters sit at roughly +8 or higher, the top 25% around +4; the bottom 10% near −9 or lower. A handful of elite seasons reach +15 to +23. Ratings are regularized toward average, so read them as tiers, not exact decimals.
- What's modeled:Shot rates, not goals.Shots are far less noisy than goals over a season, so the ratings stabilize faster and better isolate on-ice impact. They don't credit finishing or goaltending.
- Eligibility:The default Eligible players only filter keeps skaters with at least 200 weighted minutesof 5-on-5 exposure. Weighted minutes measure how much independent information a skater's shifts give the model — a player who always plays with the same linemates can log far fewer weighted minutes than raw ice time, so heavy-minute players occasionally fall below the bar. Below-bar players are kept but flagged (†), since their ratings rest on little information and are noisy.
Reference: 2024-25 distribution. Percentile levels shift modestly between seasons, so the same anchors are used as a fixed reference on every season page.