How the Concourse Rating works
The Concourse Rating turns a season of play into one score from 0 to 100. This page explains what goes into the score, how it keeps small samples from fooling it, and what it cannot see.
What the rating is
Underneath, the rating is a number of goals. The model estimates how many goals a skater added or cost his team compared with a baseline. We call that total value, and player pages show it as a plus or minus number of goals against the baseline.
The 0 to 100 score is a percentile of that total. It places a skater among all qualified skaters, forwards and defensemen together. A score of 80 means the skater’s estimated value is higher than 80 percent of the comparison group. It does not mean the skater is 80 percent effective.
Total value is offense value plus defense value plus net penalties. The Offense and Defense badges are separate percentiles inside the skater’s own position group, forwards against forwards and defensemen against defensemen. Net penalties count in the overall score and in neither badge.
The components
The overall score adds up seven components. Each one is measured in goals.
- EV offense
- The attacking impact the team gets with the skater on the ice at 5v5, estimated with RAPM and multiplied by 5v5 minutes to turn it into goals.
- EV defense
- The same idea for chances against. Fewer expected goals against with the skater on the ice counts as positive value.
- Individual
- The skater’s own 5v5 chance creation compared with the position average. It combines individual expected goals from shots, primary assists worth 0.7 goals each and secondary assists worth 0.3 goals each.
- Finishing
- Goals scored beyond what the quality of the skater’s shots predicted. A small number of shots pulls this component toward zero.
- Power play
- Power-play value compared with a lower-usage group at the same position. It needs at least 40 power-play minutes, and it counts at 25 percent in the rating so that on-ice results and individual production are not counted twice.
- Penalty kill
- How well the skater’s team prevents chances on the penalty kill, compared with the position average. It needs at least 15 penalty-kill minutes, and it counts at 50 percent in the rating.
- Penalties
- Penalties drawn minus penalties taken, compared with the position average. Each extra penalty drawn is worth 0.17 goals, and each extra penalty taken costs 0.17 goals.
A dash on a component tile means that component does not apply, for example because the skater has no qualifying power-play time. A dash is not a zero and it is not a bad score.
Why each stat is shrunk by how reliable it is
A hot month is not a skill. Before a stat counts, the model pulls the skater’s observed rate toward a baseline, and it pulls harder when there is less data. The weight on the observed rate is exposure divided by exposure plus a constant for that stat.
Each constant is a halfway point: the amount of exposure at which the observed rate and the baseline get equal weight. For example, at 500 5v5 minutes a skater’s individual expected-goal rate gets half the weight, which is why shot-based stats become trustworthy sooner than secondary assists. A smaller constant means the stat is believed sooner.
The baseline is the position average for most stats, zero for the 5v5 on-ice estimates, and a lower-usage group at the same position for the power play. Finishing is adjusted by shots on goal instead of minutes. The halfway points are below. They are minutes, except finishing, which counts shots on goal.
| Input | Sample | Forwards | Defensemen |
|---|---|---|---|
| 5v5 individual xG | 5v5 minutes | 500 | 500 |
| 5v5 primary assists | 5v5 minutes | 1,400 | 2,300 |
| 5v5 secondary assists | 5v5 minutes | 3,700 | 7,400 |
| Finishing | Shots on goal | 450 | 1,000 |
| Penalties drawn / taken | All-situation minutes | 1,400 | 1,400 |
| 5v5 on-ice offense | 5v5 minutes | 900 | 1,400 |
| 5v5 on-ice defense | 5v5 minutes | 1,300 | 1,900 |
| PP on-ice offense | PP minutes | 130 | 200 |
| PP individual xG | PP minutes | 160 | 225 |
| PP primary assists | PP minutes | 180 | 1,500 |
| PP secondary assists | PP minutes | 270 | 1,500 |
| PK on-ice defense | PK minutes | 280 | 1,000 |
These constants set how fast a stat is believed. They are not minimum-time rules for being rated.
What RAPM adds
Raw on-ice numbers give a skater credit or blame for everything that happens while he is out there, including what his linemates and opponents do. RAPM, short for regularized adjusted plus-minus, estimates on-ice expected goals for and against after accounting for the teammates and opponents a skater shared the ice with over many shifts.
The rating uses RAPM at 5v5 only. Those estimates are pulled toward zero according to ice time, using the halfway points in the table above, and then multiplied by 5v5 minutes to give EV offense and EV defense. When the 5v5 estimate is missing or pooled, the tile shows “limited 5v5” instead of a score.
Power play and penalty kill do not use special-teams RAPM. Those components use on-ice rates and individual production, because a single season cannot reliably separate players who share every shift on a unit.
How goalies are rated
Goalies have their own rating on a separate scale, and they are compared only with other goalies. Goalies do not receive the skater Offense and Defense badges.
The goalie rating is built on goals saved above expected, which measures how many goals a goalie prevented compared with what the quality of the shots faced predicted. That total is adjusted for how much the goalie has played: it is multiplied by the goalie’s minutes and divided by those minutes plus 4,500 goalie minutes. Every goalie with at least one game is ranked on the adjusted value, and the rating is the goalie’s percentile in that ranking, from 0 to 100. Save percentage, high-danger saves, NHL Edge context and earlier seasons are not part of it.
What it can’t see
A rating is an estimate from recorded data, and it has limits.
- Only the seven components count
- Anything that does not show up in those components is not in the score.
- No tracking or contract data
- NHL Edge tracking data and cap hits are shown on some pages, but neither is part of any component.
- Special teams are simpler
- Power play and penalty kill use on-ice rates and individual production, not isolated impact, so they are less adjusted for teammates than 5v5.
- A percentile is relative
- The score compares a skater with the other skaters in the same season. It is not a measure that carries across seasons.
- The source has limits
- We calculate the rating from normalized NHL game and season records. The source does not disclose every upstream collection detail.
Current season only
Ratings use games from the current season only. Last season’s results are not blended in, even early in the year.
The one exception is 5v5 on-ice impact (RAPM), which starts from a player’s previous seasons and moves as this season’s games pile up.
Early in a season the samples are small, so every stat is still shrunk by its reliability, as described above. That keeps one hot week from moving a skater far.
Boards list only players with at least a minimum number of games. The minimum is 40 percent of the league-average team games played, rounded down, and never fewer than 1 or more than 25. Each board shows the minimum in the form “Min. X GP”.
A player below that minimum can still be reached through a direct link or search. That player carries a “Small sample” tag, and the number should be read with caution. Players on the boards are not tagged simply for low minutes. Pages also show an early-season note with the average number of games teams have played so far.