Stats Glossary

Estimated total skater value against qualified league peers.

A percentile of total estimated goals relative to model baselines, combining 5v5 on-ice offense, 5v5 on-ice defense, individual 5v5 creation, finishing, power play, penalty kill, and net penalties. The overall comparison includes both forwards and defensemen. Baselines vary by component: for example, individual creation uses position averages, while power play uses a replacement-level group.

Note: Ratings use games from the current season only. A small sample is tagged and every stat is shrunk by its reliability, but last season’s results are not blended in. Goalie ratings use a separate scale.

Offensive value against other forwards or defensemen.

A percentile of offensive goal value within the player’s position group. Offensive value adds 5v5 on-ice offense, 5v5 individual creation, finishing, and power-play value; net penalties are not in this badge. Qualified players have at least 200 5v5 minutes in the regular season or 60 in the playoffs.

Note: A larger percentile means more estimated value, not a percentage of offense created. Players below a board’s minimum games carry a “Small sample” tag.

Defensive value against other forwards or defensemen.

A percentile of defensive goal value within the player’s position group. Defensive value adds 5v5 on-ice defense and penalty-kill value; net penalties are not in this badge. Lower expected goals against improve the defensive component. Qualified players have at least 200 5v5 minutes in the regular season or 60 in the playoffs.

Note: Players below a board’s minimum games carry a “Small sample” tag. Goalies do not receive these skater badges.

Estimated attacking impact with this skater on the ice at 5v5.

The player’s 5v5 on-ice expected-goals-for RAPM rate is pulled toward zero based on 5v5 minutes, then multiplied by 5v5 minutes to estimate goals. The isolated on-ice rate adjusts for teammates and opponents. If the required on-ice estimate is missing or pooled, its tile has no percentile and shows “limited 5v5”; it is not a poor score.

Estimated defensive impact with this skater on the ice at 5v5.

The player’s 5v5 on-ice expected-goals-against RAPM rate is pulled toward zero based on 5v5 minutes, multiplied by those minutes, then sign-reversed so suppressing chances is positive. If the required on-ice estimate is missing or pooled, its tile has no percentile and shows “limited 5v5”; it is not a poor score.

Own chance creation above the position average at 5v5.

Combine individual expected goals from 5v5 shots, 5v5 primary assists worth 0.7 estimated goals each, and 5v5 secondary assists worth 0.3 each. Each rate is pulled toward its position average according to its reliability before the position-average rate is subtracted and the difference is multiplied by 5v5 ice time.

Goals scored beyond what the player’s shots predicted.

Start with all-situation goals minus individual expected goals, then multiply by shots on goal ÷ (shots on goal + the position-specific reliability constant). Positive means more goals than expected; a small shot sample pulls this component toward zero.

Power-play value beyond a same-position replacement group.

Compare on-ice PP expected goals for and individual PP expected goals, primary assists (× 0.7), and secondary assists (× 0.3) with a lower-usage PP group at the same position. The reliability-adjusted rate differences are multiplied by PP minutes and counted at 25% in the rating to reduce overlap between on-ice and individual creation. At least 40 PP minutes are required; otherwise the tile shows “no PP time” instead of a percentile. Applicable percentiles compare same-position PP players.

PK chance prevention against same-position peers.

Compare the player’s on-ice PK expected-goals-against rate with the position-average PK rate; pull the player rate toward that average based on PK minutes. The difference is multiplied by PK time and counted at 50% in the rating. At least 15 PK minutes are required; otherwise the tile shows “no PK time” instead of a percentile. Applicable percentiles compare same-position PK players.

Penalties drawn and avoided versus position peers.

Drawn and taken penalty rates are each pulled toward their position averages using all-situation ice time. Their differences from those averages are multiplied by minutes and valued at +0.17 estimated goals per extra penalty drawn or −0.17 per extra penalty taken. Net penalties are a separate component tile and count in Overall, not in the Offense or Defense badges.

Short hot streaks count less than repeatable performance.

Most input rates are pulled toward their position average by exposure ÷ (exposure + the constant below). For 5v5 on-ice RAPM the target is zero; finishing has a separate shots-on-goal adjustment. Smaller constants mean the observed result gains weight sooner, so individual expected goals gain weight sooner than secondary assists. These constants are halfway points, not minimum-time eligibility rules.

PP value compared with a lower-usage player group.

For each position, the PP replacement group is the lower half by PP ice time of players with at least 40 PP minutes and 200 5v5 minutes in the season rows. Their on-ice expected-goal rate and individual expected-goal and assist rates form the comparison baseline. The other rating components use their own position averages or zero-impact references; they are not all replacement-level comparisons.

A small sample can swing quickly, so read it with caution.

Ratings use games from the current season only. Boards list players with at least a minimum number of games: 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 as “Min. X GP”. A player below that minimum can still be reached through a direct link or search and carries a “Small sample” tag. Every stat is still shrunk by its reliability, so one hot week moves a rating only a little.

A dash means no applicable score, not a bad score.

The component percentile is absent when the player has no qualifying PP time (“no PP time”), no qualifying PK time (“no PK time”), or no usable 5v5 on-ice estimate (“limited 5v5”). Ineligible components do not become an automatic 0 or 50 on the tile. Separately, missing input values can be recorded as missing and contribute zero to a goal-value sum; a dash does not describe every missing source field.

Estimated 5v5 impact beyond teammates and opponents.

Regularized adjusted plus-minus estimates on-ice expected goals for and against after accounting for teammates and opponents over many on-ice shifts. The rating uses the available 5v5 estimates, shrunk toward zero based on ice time. Power-play and penalty-kill values use on-ice rates and individual production, not special-teams RAPM. On-ice impact now uses previous seasons as a starting point, and this season’s games move it as they pile up.

A single number for how strong a team is right now, so teams can be ranked and compared.

A team rating that goes up when the team wins, more when it beats a good team or wins big, and drifts back toward average over the summer. A loss moves it down by the same logic, so beating a weak team barely changes it.

Note: Elo ranks how well a team has played, not who has the best roster on paper. The Power Rankings page shows rank movement over the last 7 days.

Turns a standings race into a chance, so one more win or loss is easy to weigh.

The share of 20,000 simulated rest-of-seasons in which the team makes the playoffs. Each simulation plays out every remaining game using the teams’ Elo ratings, then applies the real standings rules. The same simulations give each team’s chance of winning its division, reaching each round and winning the Cup.

Note: Head-to-head and goal-differential tiebreakers are not simulated. Odds under 1% show as "<1%" and over 99% as ">99%"; a team only shows 0% or 100% once its points make that certain. The simulation is rerun every night.

Shows how much on-ice value a team gets for each million dollars of cap space a player takes up.

Take the goals of value the Concourse Rating credits a skater with this season, then divide by the player’s cap hit in millions. A skater worth 4.0 goals on a $2.0M cap hit returns +2.0 goals per $1M. The percentile compares that figure with every other skater who has enough games to qualify and a cap hit.

Note: It is a current-season figure, so small samples move fast. Skaters with a small sample or no cap hit on file show a dash instead of a number, and a dash is never a zero. Entry-level and minimum contracts tend to rank well because the cap hit is small. The least-value board only considers cap hits of $5.00M or more.

Shot-attempt share shows which team drove more play on ice.

The worker supplies on-ice shot attempts for (CF) and against (CA), including shots on goal, misses, goals, and attempts that are blocked. The site calculates CF% as CF ÷ (CF + CA).

On-ice shot-attempt share compared with off-ice share.

The player's on-ice CF% minus the team's off-ice CF%. The site derives the off-ice share from team totals after subtracting the player's on-ice attempts.

Share of unblocked attempts while the player is on ice.

The worker supplies unblocked attempts for (FF) and against (FA): goals, shots on goal, and misses, but not attempts blocked before reaching the net. FF% is FF ÷ (FF + FA).

On-ice unblocked-attempt share compared with off-ice share.

The player's on-ice FF% minus the team's off-ice FF%, with the off-ice share derived from team totals after subtracting the player's on-ice unblocked attempts.

Context for unusually high or low on-ice results.

The team's on-ice shooting percentage plus its on-ice save percentage while the player is on the ice. The site calculates shooting percentage as goals for ÷ shots for × 100 and save percentage as (shots against − goals against) ÷ shots against × 100, then adds them.

Share of counted shifts begun in the attacking zone.

The worker-supplied offensive-zone starts divided by offensive-zone starts plus defensive-zone starts. Neutral-zone starts and on-the-fly changes are not in this denominator.

Share of counted shifts ending in the attacking zone.

The worker-supplied offensive-zone finishes divided by offensive-zone finishes plus defensive-zone finishes. It is a share of counted zone finishes, not a measure of total seconds spent in each zone.

Chances weighted by how likely they were to become goals.

The worker supplies expected-goal values: probability-based estimates of how often attempts with the model’s characteristics become goals. The site receives individual xG for the shooter and on-ice xGF/xGA totals for attempts taken by either team; the worker contract does not expose the model’s feature list here.

Rate of quality chances for and against while on ice.

On-ice expected goals for (xGF/60) and against (xGA/60) are expected-goal totals scaled to 60 minutes of the same game situation. They describe team chances during the player’s shifts, not the individual’s shots. The PP and PK components use their respective on-ice rates; the 5v5 isolated-impact input is a separate RAPM estimate.

Quality-weighted chances from the player’s own shots.

The worker supplies the sum of expected-goal probabilities on the player’s own attempts. The rating uses a reliability-adjusted 5v5 ixG rate in EV Individual, an all-situations ixG total in Finishing, and a PP ixG rate in the power-play component.

Direct and earlier credited setups for a goal.

A primary assist is the final assist before a goal; a secondary assist is the other credited assist. EV Individual and Power Play value the primary-assist rate at 0.7 estimated goals and the secondary-assist rate at 0.3, then adjust each rate for reliability against a position baseline. A1/60 on the card is a separate rate display.

Share of quality chances created by the player’s on-ice team.

The worker supplies the player's on-ice expected goals for and against totals. The site calculates the share as xGF ÷ (xGF + xGA) and displays it as a percentage.

Actual goals minus the player’s expected goals.

A skater's goals minus individual expected goals: goals − ixG.

Note: This raw difference is not the Finishing rating component; that component also applies a shots-on-goal reliability adjustment.

Share of high-danger chances created while on ice.

The worker supplies probability-defined high-danger attempts for and against. The site calculates HDCF% as high-danger attempts for ÷ (high-danger attempts for + high-danger attempts against). The worker contract identifies these as probability-based but does not expose the cutoff or model features.

A game summary, separate from the current skater rating.

A weighted production total: goals × 0.75, primary assists × 0.70, secondary assists × 0.55, shots on goal × 0.075, blocks × 0.05, penalties drawn minus taken × 0.15, faceoff wins minus losses × 0.01, 5-on-5 Corsi for minus against × 0.05, and 5-on-5 goals for minus against × 0.15.

How often the player earned a point on an on-ice goal.

The player's goals plus primary and secondary assists, divided by the team's on-ice goals for, then multiplied by 100.

Compare event frequency across players with different ice time.

A counting total divided by ice time in seconds and multiplied by 3,600. For example, 10 goals in 1,000 minutes is 0.60 goals per 60.

Note: Always compare the same situation. Goals/60 and A1/60 on the card are rate tiles, not the formulas used for the Offense or Defense badges.

Minutes and deployment explain how many chances a player had to add value.

Time on ice is tracked by situation: 5v5, power play, penalty kill, and all situations. The rating uses situation-specific minutes for its components and all-situation minutes for penalties. More time is opportunity, not proof of better play.

Speed, shooting, and zone-time context from tracking.

Player pages and cards can show NHL Edge fields such as skating speed, shot speed, and offensive-zone time when available. These are separate from the worker-provided 5v5 and special-teams inputs described above. A displayed Edge field is not itself one of the seven skater value components.

Contract cost alongside hockey performance, not part of the rating.

The player’s current-season cap hit comes from CapWages contract data when available. It is a salary-cap figure, not estimated goals or a rating input; contract figures may be unavailable for some players.

Goals a goalie saved beyond what the shots faced predicted, ranked among goalies.

Goals Saved Above Expected (GSAx) this season, adjusted for how much the goalie has played: GSAx multiplied by minutes and divided by minutes plus 4,500. Every goalie with at least one game is ranked on that 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 the rating.

The share of shots on goal the goalie stopped.

The worker supplies all-situations shots faced and goals allowed. The site uses (shots faced − goals allowed) ÷ shots faced.

Goals prevented beyond the quality of shots faced.

The worker's expected goals faced minus actual goals allowed. Positive means the goalie allowed fewer goals than the model expected; negative means more.

Stop rate against shots the worker classifies as dangerous.

The worker's probability-defined high-danger attempts faced minus high-danger goals allowed, divided by high-danger attempts faced. The contract does not expose the model cutoff used to classify an attempt as high danger.

A shot-quality baseline for actual save percentage.

A shot-quality baseline calculated as 1 − (expected goals faced ÷ shots faced), when both values are available. It represents the save percentage implied by the worker’s expected-goals total.

Shows how often a goalie faces a dangerous second chance after an initial shot.

Rebound shots faced divided by all shots on goal faced, multiplied by 100, using shots attributed to the goalie in the worker’s shot records. A lower rate receives a higher rebound-control percentile.

A high or low result from a short appearance sample should not look as certain as a full-season result.

Goalie card component percentiles compare against goalies with at least 20 games played or 1,000 minutes in the season. Goalies below both thresholds have a limited sample, so their component percentiles are drawn toward a neutral 50. There is no prior-season carry-over in goalie card percentiles.

Puts the goalie’s workload and usage in the context of their team.

A goalie with fewer than 10 games played is labeled Limited. Otherwise, a goalie with at least 55% of team starts is a Starter; 35% to less than 55% is a 1B; below 35% is a Backup.