Goals For Percentage in Hockey (GF%)

Goal share is plus-minus turned into a percentage. The NHL's glossary calls plus-minus "hockey's first enhanced stat" and dates it as an official statistic to 1959-60. It's a player's goal differential while he's on the ice, with power-play goals left out. Plus-minus counts the gap. GF% divides a team's goals by all the goals, which puts a fourth-liner and a top-pair defenseman on the same scale.

The league files the share as goals for percentage, with numbers back to 2009-10, and its glossary describes the skater version as generally five-on-five. The entry says to compare it with SAT% and USAT% (Corsi and Fenwick by their NHL names). The even-strength version's entry adds the warning that goal samples build up much more slowly than shot samples.

Natural Stat Trick publishes GF% for every team and skater, MoneyPuck carries on-ice goals for and against in its data files, and the NHL's team percentages report lists a five-on-five goals for percentage for every club. They're all built from the same thing, the goals logged in the league's play-by-play with the skaters who were on the ice for each one.

Count the goals your team scored and the goals it allowed in the minutes that matter, then divide your goals by the total. At five-on-five that leaves out every goal scored in other game states (power plays, shorthanded play, empty nets, three-on-three overtime), about a third of the league's scoring last season. These are the inputs:

• Goals for at five-on-five
• Goals against at five-on-five
• For a player, only the goals scored while he was on the ice, by either team
• Five-on-five ice time, for per-60 rates

Here's a worked example, with invented round numbers. A winger is on the ice for 40 Detroit goals and 30 opposing goals at five-on-five over a season. His GF% is 57.1%, and his five-on-five plus-minus on those goals is plus-10. A teammate on for half as many goals each way has the same share on half the sample, which is why the share travels better than the count.

Every skater on the ice gets the goal in his column whether he touched the puck or not, the same rule plus-minus uses. That's why a player's GF% says as much about his linemates and the goalie behind him as it does about him.

GF% measures results. It's the scoreboard, cut down to the minutes and the game state you choose, and over a full season it's the share that lines up with the standings. Among the 32 teams last season, five-on-five GF% correlated at 0.86 with points percentage (Corsi share managed 0.57).

Colorado led the league with 62.8% of the five-on-five goals in its games, and Vancouver finished last at 38.4%. Players spread far wider than they do on shot shares, because goals are rare. Among the 598 skaters with 500 five-on-five minutes, the median on-ice GF% was 50.0%, the top tenth cleared 59.5% and the bottom tenth sat at or below 39.3%, a spread nearly twice as wide as Corsi's. Alex DeBrincat's 56.2% topped Detroit's regulars.

Does a good GF% predict more of the same? Less than you'd think, and less early. In this site's split-season test (96 team-seasons over the last three years), a team's GF% through its first 20 games correlated at 0.37 with its GF% the rest of the way. Expected goals managed 0.51 and score-adjusted Corsi 0.42. By the halfway point goal share had caught up, and first-half GF% was the best single forecast of second-half GF% in the sample at 0.56, edging expected goals at 0.52. Goals get there. They just need half a season to do it.

Checking results against process

Detroit's 2025-26 season is the clean example. The Wings took 48.8% of the five-on-five attempts and the same share of the expected goals, and came away with 45.5% of the goals. Had their goal share matched their chance share, the five-on-five goals in their games would've split about 152 to 160. They split 142 to 170. That's ten goals each way, and most of it traces to the Wings' shooting (the PDO page breaks it down).

Reading a player's on-ice results

Put a player's GF% next to his expected-goals share and the gap tells you how much finishing and goaltending did for him, since the expected-goals side has already priced every chance he was on the ice for at both ends. Nathan MacKinnon led the league at 70.4% on a 57.6% share of the expected goals (both at five-on-five). Marco Kasper went the other way, on for 38.2% of the goals and 50.3% of the expected goals, a goal share near the bottom of the league while his chance share sat at even.

Choosing the game state

Five-on-five GF% is the default because power-play time varies so much from team to team, but it can hide a special-teams story, and a team that wins or loses its power-play minutes by a wide margin looks different once those goals are added back in. Detroit's all-situations goal share was 48.5%, three points better than at five-on-five, because the Wings outscored opponents 54 to 47 on the power play and penalty kill combined.

Replacing plus-minus

Why not just use plus-minus? Because it mixes in shorthanded and empty-net goals and says nothing about how much a player played, while five-on-five GF% drops the special-teams noise and, as a share, doesn't reward a player just for piling up minutes. Two skaters who are each plus-10 can carry different goal shares, depending on how many goals were scored around them, and the share is the one you can compare across a roster.

Goals are rare

Detroit's whole five-on-five season came down to 312 goals, against 7,734 shot attempts. Each goal moves GF% roughly 25 times as far as one attempt moves Corsi.

Goalies and shooting luck ride along

A hot goalie or a cold stretch of finishing swings GF% without any change in how a team plays. PDO measures that part. It's shooting percentage plus save percentage, and it drifts back toward 100 over time.

A player's number is partly his team's

Among the 598 skaters with 500 five-on-five minutes last season, on-ice GF% correlated at 0.53 with the team's GF%, since linemates and the goalie behind a player both leak into his on-ice goals.

It repeats poorly from year to year

Among the 501 skaters with 500 minutes in both of the last two seasons, on-ice GF% correlated at 0.30 year over year, against 0.63 for on-ice CF% over the same span.