Game Grade describes one performance. Season rank describes this season’s results relative to workload. Rolling Radar carries recent history across seasons.
Five parts make up one performance score: Offense (25%), Defense (25%), Finishing (10%), Assisting (20%), and Chance Creation (20%). Offense and Defense measure on-ice xGF and xGA against what a player’s deployment would suggest. Chance Creation is the player’s own xG; Finishing is goals minus individual xG; Assisting credits primary assists plus half of secondary assists, each against its fixed baseline.
Each part uses a fixed scale for the position. The weighted parts add up to the score shown in the game explanation — they’re not points out of 100. That score is scaled for a single game, then turned into a percentile against historical players at the same position.
The grade covers all rated strength states with both goalies present, excluding empty-net play. A 5v5 table can therefore show different totals. At 5v5, on-ice expectations include teammate, opponent, score, venue, time, recent faceoff-zone, and rest context. These remain shared results, not a player’s isolated impact.
GSAx measures how many goals a goalie saved above what the shots were expected to produce. The grade shows how much that result stands out, given how much those shots normally vary.
Grade index = stopping residual / expected standard deviation of shots faced
| Ten shots, each with… | 10% chance | 50% chance |
|---|---|---|
| Expected / allowed goals | 1 / 0 | 5 / 4 |
| GSAx | +1 | +1 |
| Variance | 0.90 | 2.50 |
| Expected variation (√variance) | 0.95 goals | 1.58 goals |
| Game index | +1.05 | +0.63 |
The first result stands out more. A shot near 50% is the hardest to predict; very low or very high probabilities are more predictable. With the same GSAx, more variance lowers the index. Stopping harder shots also earns more GSAx, which can offset that.
This measures verified 5v5 shot-stopping only, not a goalie’s overall play. The variance math treats each shot as independent, so it won’t capture every real-world link between shots. It describes how much shot results vary — not how confident we are in the goalie’s ability. It needs verified shots and variance to compute; if inputs are missing, no grade is shown.
Save percentage, wins, shutouts, and danger-tier stats don’t add bonus points to the grade. This explanation stays fixed no matter what situation or adjustment is selected elsewhere. Its verified model totals can differ from the xGA and GSAx shown in a game breakdown.
Skaters and goalies use separate fixed benchmarks and the same letter boundaries. A percentile compares the game index with historical appearances; it is not a probability or ability-confidence score. Other players in the current game cannot change the grade.
A+ ≥88 · A ≥78 · A− ≥68 · B+ ≥58 · B ≥50 · B− ≥42 · C+ ≥35 · C ≥28 · C− ≥22 · D ≥15 · F <15
Skaters’ scores are added up per 60 rated minutes. Goalies’ verified 5v5 GSAx is added up per 100 verified shots. A fixed, position-specific formula places each rate on a 0–100 scale, where 50 is the fixed baseline. Game letters and percentiles are never averaged in.
Season rank and percentile compare unrounded rates within the current position group. A player can get a rating after a single valid appearance — check games, minutes or shots, and coverage. Early results can move a lot. Rolling Radar is separate from this rating.