Analytics, 2010-11

The regular season and the playoffs are kept apart, and the Games box above picks one. The numbers below start in 2010-11, the first season the NHL recorded every shot attempt and where it was taken from. Five-on-five means both teams at full strength with their goalies in. It is most of a game and the fairest place to compare teams and players, so power plays do not skew the numbers.

Player rating

Goals above replacement

One number for a player's season: the goals he added to his team's goal difference, compared with a spare-part player given the same ice time, the kind a team can call up for nothing. It is a total, so playing a full season counts for more than playing half of one. Among regulars most players fall between -5 and +10, and the best seasons are near +30.

Wins above replacement

The same in wins. About six goals of goal difference are worth one win in the standings.

The five parts

Even-strength offence is his own goals, assists and chances at five-on-five, and how much better the team's chances are with him on the ice than with him off it. Even-strength defence is the same comparison for chances against, with his blocked shots and takeaways. Power play is a fifth of what his unit did above the league rate, tilted towards whoever scored or made the last pass. Penalty kill is a quarter of what his unit gave up above the rate. Penalties is the minors he drew less the minors he took, at about a sixth of a goal each. Each part is zero for an average skater.

Goalies

A goalie's rating is his goals saved above expected, plus the same kind of allowance for workload. Only about a third of a goalie's season figure reflects the goalie, so one year should be read lightly.

What it cannot see

It does not pay for scoring more than the chances were worth, so pure finishers rate lower than their goal totals suggest. It does not know who a player was sent out against, so shutdown players are under-rated. A star on a deep team is compared with good teammates and loses a little by it. A full season is about two-thirds the player and one-third chance, and under forty games it is more chance than player.

Shot attempts

Corsi (CF, CA, CF%)

Every shot attempt: shots on goal, missed shots and blocked shots. CF is attempts for a team and CA attempts against it. CF% is the share that were for. Above 50% means a team, or the team while a player is on the ice, has the puck more than its opponents. It stands in for possession, which nobody records directly.

Fenwick (FF%)

The same without blocked shots, on the view that a blocked shot was never much of a chance.

SF%, GF%

The share of shots on goal, and of goals, that were for.

Sh%, Sv%, PDO

Sh% is the share of a team's shots on goal that went in, and Sv% the share of the other team's that were stopped. PDO is the two added together. Across the league it comes to 100. A team or player far above 100 is getting the bounces and usually cools off; one far below is usually due to recover.

iCF, attempts taken

The shot attempts a player took himself.

Expected goals

Expected goals (xG)

What a shot was worth. Every unblocked attempt gets the chance that a shot like it goes in, judged by where it was taken from, what kind of shot it was, whether it was a rebound and the score. A shot from the slot is worth about 0.20 and one from the point about 0.015. Add them up and you have the goals a team or player would expect from the chances they had. Blocked shots count for nothing, because they are recorded where they were blocked.

Seasons before 2023-24

The NHL timed events less exactly before 2023-24, and differently from rink to rink, so expected goals for those seasons are rougher. They still add up to the goals actually scored each season, and they are still a fair guide to individual shooters and goalies. For teams before 2022-23 they are no better a guide than Corsi.

xGF%

The share of the expected goals at five-on-five that were for. Like CF%, but each attempt counts for what it was worth, so a team that gets to the slot is rated above one that shoots from the boards. For a skater it is the share while he was on the ice.

Chances created (xG per 60)

The expected goals of a skater's own shots for every 60 minutes he plays: how often he gets a chance.

Shot quality (xG per attempt)

The average chance of one of his unblocked attempts going in: how good each chance is.

Finishing (goals above expected)

Goals minus expected goals. Over one season it is mostly luck, and it takes several seasons before it says much about a shooter. Shots at an empty net are left out, so an easy goal does not count as finishing.

A player against his team

CF% rel, xGF% rel

A player's share while he is on the ice, compared with his team's share in the same games when he is off it. It separates a good player on a poor team from a passenger on a strong one. +5 means the team does five points better with him out there.

Lines and pairings

Forward lines and defence pairs

On team and player pages: the three forwards, or the two defencemen, who were on the ice together at five-on-five, how long for, and how the team did in that time. It is worked out from the shift charts a second at a time. A line has to be out together for twenty seconds in a game to count, so players crossing during a change are not mistaken for a line.

Goalies

Goals saved above expected

The expected goals of the shots a goalie faced, minus the goals he let in. It judges him on how hard his shots were and not just how many he stopped, so a goalie behind a leaky defence gets credit that save percentage does not give him. Only shots taken while he was in net count. Each season is set so the league's goalies as a whole come out at zero, which makes it a measure against his peers.

Per 60

The same for every 60 minutes he played, so a backup can be compared with a starter.

Shot difficulty (xG per attempt faced)

The average chance of a shot against him going in. A high number says more about his team than about him.

Expected save %

The save percentage an ordinary goalie would have had on the same shots, to set beside his own.

Contracts and value

Cap hit

What a player counts against his team's salary cap in a season. It is the average of his contract over its years, so it can differ from what he is paid in any one of them, and it is the same whether he plays every game or misses half the season.

Retained salary

A team trading a player may keep paying part of his contract. He is still shown with his whole cap hit, and judged against it, but his new team is charged only its share: that share is what counts in a team's value and on its roster (marked with a star), and what he costs in Be a GM. The league's rules: a team may keep at most half of a contract, may be keeping part of no more than three contracts at once, and may not have more than 15% of the cap ceiling tied up this way. A contract can have salary kept by at most two teams, and from 2026-27 the second cannot happen within 75 days of the regular season after the first, which ends trades that passed a player through a third team to cut his cost twice.

Dead cap

What a team counts against the cap for players who are no longer on it: the salary it kept paying for a player it traded, and what is left of a contract it bought out. Only the cases gathered so far are listed, on the team's page and in Be a GM.

Buyout

A team may end a contract early by paying the player two-thirds of the salary he is still owed, or one-third if he is under 26, spread over twice the years that were left. Signing bonuses are paid in full on top. What counts against the cap each year is his cap hit less what the team saves in salary that year, so the charge can be large in a year his salary was low, runs twice as long as the contract would have, and can even be a small credit. Buyouts are made in a window from mid-June (or two days after the Stanley Cup is won, if later) to June 30.

Where the contracts come from

The NHL does not publish contracts, so they do not come with the games. They are gathered from public collections of contract data and from news reports of signings. Recent seasons are nearly complete. Further back, mostly the contracts of players who were still playing in 2024 are known, so an older season can be missing most of its players, and the page says so when it is.

Cap dollars per point, per goal above replacement

The cap hit divided by what the player produced in the regular season. Lower is better value. In a season still being played these start very high and fall as the games add up.

The going rate

What teams as a whole paid for one goal above replacement that season: the cap space they spent above the cheapest contract, divided by the goals above replacement they got for it. Skaters and goalies are priced apart, because a goalie's goals above replacement run much higher and each costs less.

Worth at the going rate

What a season like his cost on average: the cheapest contract plus his goals above replacement at the going rate. A player who rated below a spare-part player can come out worth less than nothing.

Value

Worth less cap hit: how much more, or less, he gave than he cost. Players on their first contract are paid far below the going rate, so they top the lists. Only players with enough games to be ranked get a value, and it is for the regular season only.

A team's value

The same, added up over its ranked players whose contract is known. A player who changed teams is shared out by the games he played for each.

Reading the pages

Who is ranked

A call-up with one good night would top every list, so lists of rates and shares only include players who played a fair amount: at least a quarter as much as the busiest player that season. Totals, such as most expected goals, include everyone.

Ranking bars

On a player's page, a bar filled to the 90th means he is ahead of 90% of the regulars at his position. Forwards are ranked against forwards, defencemen against defencemen and goalies against goalies.

Coloured numbers

On player and team pages a number is coloured by its percentile, the share of the others it is ahead of: dark green from the 80th up, green from the 60th, amber for the middle, pale red from the 40th down and dark red from the 20th down. Hover over the number for its exact percentile.

Who it is ranked against depends on the number. A skater's on-ice shares are set against every skater who played enough to be ranked, and his rating against the ranked players at his position. A goalie's numbers are set against the regular goalies, a team's against the other teams, and a line or pair against the lines or pairs that played at least a tenth as long together as the busiest one.

Numbers that are neither good nor bad on their own stay plain: PDO, which is mostly luck, and counts that simply follow ice time, such as attempts taken or goals against. A rating is also left plain until a player has enough games to be ranked.

The radar on the compare page

Each spoke is one measure, running from the lowest regular at the centre to the league leader at the rim. The gap between two players on a spoke is the real gap in their numbers, and the dashed outline is the typical player.

Race charts

Stepped lines that add up through a game or a season. On a game page each team's line climbs with every chance, with a dot for each goal. On a player's page his goals race his expected goals, and on a goalie's page the goals he allowed race the goals he was expected to allow.

Shot maps

The attacking half of the rink with the net on the right. A map with many shots opens as a heat map: darker red means more shots from that area, and each dot is a goal. The button under the key switches to a dot for every shot, sized by how good the chance was. Clicking Goals, Saved, Missed or Blocked in the key shows only those.