How Hockey ChangedMinutes

The NHL Bench Grew From Eight Skaters a Night to 18 in Jumps

In 1923 the Stanley Cup champions used nine skaters all season, and one substitute filled in for all five regulars in the Final. Today every team dresses 18. Two of the jumps came in the same season as a published change in the dress limit (the most skaters a team may put in uniform for a game); the step to 18 lines up only with a table of limits on a hockey-history website that names no author. The 60-minute player was gone long before anyone timed him.

Verdict: Data story

By Edgehalla ResearchPublished 11 min read

Post on X

In the 1923 Stanley Cup Final against the Edmonton Eskimos, each of Ottawa’s five regular skaters “had to go off for repairs or because of fatigue,” as NHL.com tells it. King Clancy, a 20-year-old substitute, filled in for every one of them.

At some point in those playoffs, the story goes, the Senators’ goalie, Clint Benedict, had to serve a penalty. On his way off he threw his stick to Clancy. “Here kid,” he said, “take care of this place until I get back.”

That season the Senators used nine skaters, eight a night, and won the Cup. Every NHL team now dresses 18.

More takeaways below

  • No skater comes close to 60 minutes now. A team’s busiest skater plays about 25 minutes a night. The highest season average since ice time was first published in 1997-98 is Ryan Suter’s 29:25 in 2013-14, and the most in one regular-season game is Cam Fowler’s 38:54.
Use the toolNHL line combinations todayWho is on which line and power-play unit tonight.

Ottawa won the Cup with nine men

Nine was a full roster then. Six Senators, Clancy among them, are listed in the official records as playing all 24 regular-season games. Eddie Gerard missed one, Frank Nighbor two, and Lionel Hitchman played three. Add up their games, divide by the schedule, and Ottawa dressed 8.0 skaters a night. In the whole history of the league, only the 1921-22 Hamilton Tigers (7-17-0, in wins, losses and ties) also got through a full schedule with nine. Four years earlier, in 1918-19, four Senators had played every game: Cy Denneny, Eddie Gerard, Sprague Cleghorn and Frank Nighbor.

The Clancy legend grew from there. NHL.com says he played every position during a Stanley Cup playoff game, “the only time it has ever been done.” The historian Eric Zweig went back to the Ottawa Journal. Clancy “played every position on the ice during the 1923 Stanley Cup playoffs,” Zweig wrote, but “he did NOT do so in one single game.”

The Journal’s game report of April 2, 1923, as Zweig reads it, has Clancy and Hitchman subbing on defence, “with Harry Helman taking a few turns relieving the forwards.” A column by Basil O’Meara in the same day’s Journal said Clancy “went in and played goal and tried his hand at every other position on the team.” Zweig ends “not quite as sure as I once was.”

The whole league ran that lean. Add up every skater’s games in an NHL season, divide by the games each team played, and you get how many skaters a team used on an average night. We did that for every season since 1917-18. In 1917-18 the answer was about eight. Five skaters were on the ice at a time, plus the goalie (the National Hockey Association, the NHL’s forerunner, went to six a side, goalie included, in 1911-12). About three more sat on the bench to give them a rest. The count passed 10 in 1926-27 and reached 15.6 in 1949-50. It has been almost exactly 18 in every season since 1982-83. Goalies are counted separately and are not in these numbers. Since 2017-18 we can check this count against box scores, and it matches almost exactly. The early decades are rougher, because the record keeping then was patchier.

Two of the jumps on the way up came in the same season as a published change in the dress limit. A few more line up with an unsigned table of limits, which we come to below. Others, like the climb back after the Second World War, came with no limit change on any list. In some seasons the busiest teams dressed exactly as many skaters as that table allowed. In others they stayed well under.

The bench grew in jumps; two came with a published change in the dress limit

NHL regular season, every season since 1917-18 (goalies not included)

Show data
NHL teams used 7.96 skaters a game in 1917-18, about 15.4 in the 1950s and 18.00 in 2025-26.
SeasonSkaters per team game
1917-188.0 skaters
1918-197.9 skaters
1919-208.1 skaters
1920-218.0 skaters
1921-228.5 skaters
1922-238.0 skaters
1923-248.6 skaters
1924-259.0 skaters
1925-269.9 skaters
1926-2710.1 skaters
1927-2810.0 skaters
1928-2910.2 skaters
1929-3011.8 skaters
1930-3112.5 skaters
1931-3213.0 skaters
1932-3313.3 skaters
1933-3413.5 skaters
1934-3513.6 skaters
1935-3613.7 skaters
1936-3713.8 skaters
1937-3813.8 skaters
1938-3914.2 skaters
1939-4013.6 skaters
1940-4113.6 skaters
1941-4213.6 skaters
1942-4312.9 skaters
1943-4412.2 skaters
1944-4513.0 skaters
1945-4613.8 skaters
1946-4714.6 skaters
1947-4814.6 skaters
1948-4914.8 skaters
1949-5015.6 skaters
1950-5115.7 skaters
1951-5214.7 skaters
1952-5315.0 skaters
1953-5415.3 skaters
1954-5515.3 skaters
1955-5615.4 skaters
1956-5715.3 skaters
1957-5815.5 skaters
1958-5915.8 skaters
1959-6016.1 skaters
1960-6116.0 skaters
1961-6215.5 skaters
1962-6315.6 skaters
1963-6415.7 skaters
1964-6515.9 skaters
1965-6615.7 skaters
1966-6715.8 skaters
1967-6815.4 skaters
1968-6915.5 skaters
1969-7015.7 skaters
1970-7116.6 skaters
1971-7216.3 skaters
1972-7316.2 skaters
1973-7416.4 skaters
1974-7516.4 skaters
1975-7616.5 skaters
1976-7716.7 skaters
1977-7816.8 skaters
1978-7916.8 skaters
1979-8016.9 skaters
1980-8116.9 skaters
1981-8216.9 skaters
1982-8317.9 skaters
1983-8417.9 skaters
1984-8517.9 skaters
1985-8617.9 skaters
1986-8718.0 skaters
1987-8818.0 skaters
1988-8918.0 skaters
1989-9018.0 skaters
1990-9118.0 skaters
1991-9218.0 skaters
1992-9318.0 skaters
1993-9418.0 skaters
1994-9518.0 skaters
1995-9618.0 skaters
1996-9718.0 skaters
1997-9818.0 skaters
1998-9918.0 skaters
1999-0018.0 skaters
2000-0118.0 skaters
2001-0218.0 skaters
2002-0318.0 skaters
2003-0418.0 skaters
2004-05–
2005-0618.0 skaters
2006-0718.0 skaters
2007-0818.0 skaters
2008-0918.0 skaters
2009-1018.0 skaters
2010-1118.0 skaters
2011-1218.0 skaters
2012-1318.0 skaters
2013-1418.0 skaters
2014-1518.0 skaters
2015-1618.0 skaters
2016-1718.0 skaters
2017-1818.0 skaters
2018-1918.0 skaters
2019-2018.0 skaters
2020-2118.0 skaters
2021-2218.0 skaters
2022-2318.0 skaters
2023-2418.0 skaters
2024-2518.0 skaters
2025-2618.0 skaters
  • Second World War: 1939-40 to 1945-46
  • 1922-23: Ottawa: 9 skaters all season
  • 1949-50: Limit raised to 17 skaters
  • 1951-52: Cut to 15 plus goalies
  • 1982-83: Step to about 18
  • Reference: Today’s limit: 18 skaters (18.0 skaters)
Skaters per team game: every skater’s games played added up and divided by team games played. 1917-18 leaves out two Montreal Wanderers games that were never played after the team folded but were still counted. Early records of who played may not match who dressed exactly. The 2004-05 season was cancelled.

Most of the climb came in a few bursts. The fastest ran from 1928-29, at about 10 a game, to 1931-32, at about 13. During the Second World War the count dipped to about 12 in 1943-44, the fewest since 1929-30. By 1946-47 it was back to about 15. From 1949-50 to 1969-70 it stayed between about 15 and 16. From 1970-71 to 1981-82 it sat a step higher, between about 16 and 17. In 1982-83 it jumped to almost 18.

Changing on the fly did not make the bench any bigger

The first suspect is the way players came on. Herb Gardiner joined the Montreal Canadiens at 35, in 1926-27. He played all 44 of their games, and the Manitoba Hockey Hall of Fame says “playing every minute of every game that year for the Canadiens earned him the nickname, ‘the Ironman of hockey’.” Nobody kept ice time then, so “every minute” is the Hall’s account, not a timed figure. Gardiner won the Hart Trophy as the league’s most valuable player that season.

The next season the NHL changed how players came on. The refereeing site Scouting the Refs, in its history of NHL rule changes, lists it for 1927-28: “Teams permitted to change on the fly.” From then on a coach could send a fresh skater over the boards while play went on, without waiting for a whistle. Teams went from 10.10 skaters a game to 10.04, a hair lower. Coaches got a new way to use their bench and kept it the same size.

What the NHL did change, several times, was its dress limit. The table below lists every change we could confirm in a published rule history, with the count the season before and the season of the change. Read down the last two columns: in 1949-50 the count goes up with the limit, and in 1951-52 it comes down with it. The other rows move a little or not at all.

Two limit changes move the count clearly, the same way the rule moved: up in 1949-50, down in 1951-52. Dress-limit and substitution rules from a published NHL rule history (Scouting the Refs), with skaters used per team game the season before and the season of the change.
SeasonRuleSeason beforeThat season
1925-26Dress at most 12 players from a roster of 149.039.88
1927-28Line changes on the fly allowed10.1010.04

Continues below

Rows 3–7 of 7
SeasonRuleSeason beforeThat season
1928-29Dress at least 8 and at most 12 skaters10.0410.22
1932-33Players in uniform cut to 14 from 1513.0413.27
1949-50Dress 17 players, not counting goalies14.8515.55
1951-52Cut to 15 players plus goalies15.7014.69
1965-66Two goalies must dress for every game15.8815.73

In 1949-50 the limit went up and teams used about 0.7 more skaters a game. In 1951-52 it was cut and they used about 1.0 fewer. The two-goalie rule of 1965-66 shows up on the goalie side instead. The season before, a second goalie got into a team’s game about once in 60 games. In 1965-66 it was about once in 10. That was the first regular season with a backup in uniform every night (playoff backups had to dress from 1964-65).

One row runs the wrong way. In 1932-33 the limit was cut to 14, yet teams used slightly more skaters a game, about 13. The rule history doesn’t say whether the goalie was one of the 14. If he wasn’t, teams were already under the new limit. If he was, 13.3 skaters plus a goalie comes to more than 14, and the records and the rule disagree. The list has a hole too. It says the limit came down from 15, but no earlier entry gives 15.

For years the counts ran past the published limits

If the published rules set the size of the bench, the counts should never run past them. They do, for years at a time. The last skater limit the rule history gives is 15 plus goalies, set in 1951-52. Yet the league averaged more than 15 skaters a game in every one of the 14 seasons from 1953-54 to 1966-67. In 1959-60 it reached about 16. It happened earlier too. In 1931-32 teams averaged 13.0 skaters a game against the limit of 12 set in 1928-29. In 1938-39 they averaged 14.2 against the 14 in uniform set in 1932-33. For 1982-83, when the count jumped to almost 18, the rule history lists no change at all.

A team can’t dress more skaters than the rules allow, so something in the record is wrong. Either the old records list some players in games they never dressed for, or the rule history is missing changes. Only two of its limits clearly show up in the count: 17 in 1949-50 and 15 in 1951-52.

A season-by-season table on a hockey-history website, adjustedhockey.com, points to missing changes. Nobody signed it. It cites the stats site Hockey-Reference in general and gives no source for each season. In it the limit moves far more often: 15 for 1929-30 to 1931-32, 15 again for 1938-39 to 1941-42, 16 for 1953-54, 17 from 1954-55 to 1959-60, 16 from 1960-61 to 1970-71, 17 from 1971-72, and 18 from 1982-83. One season had two limits. In 1952-53, the table says, teams could dress “16 skaters at home and 15 on the road.”

Put the website’s dates next to the counts. In 1931-32, 1938-39 and every season from 1953-54 to 1966-67, the league ran past the rule history’s limit and still came in under the website’s. The 13.0 skaters of 1931-32 sit under its 15. The 16.1 of 1959-60 sit under its 17. For those seasons the website’s dates fit the bench. The late 1940s and 1970-71 are another matter, and we come to them below. We still can’t say who made the table or where each number came from.

The jump to 18 probably came with a new limit too

The table’s last step is the 1982-83 jump to 18, and one team fits it to the skater. That September the Washington Capitals traded Ryan Walter and Rick Green to Montreal for Rod Langway, Craig Laughlin, Doug Jarvis and Brian Engblom. After that, none of their skaters played for another NHL team all season, which means they can be counted exactly. The league’s season records list each player once, with one total of games. For a traded player, that total can’t be split between his teams. But when nobody on a team played anywhere else, every game in every total belongs to that team.

So add them up. In all, 26 skaters shared the work for Washington, and their games come to 1,440 over 80 games: exactly 18 a night. Every seat was filled in the first season the website’s table allows 18. Five Capitals played all 80 games (Bobby Gould, Rod Langway, Doug Jarvis, Dennis Maruk and Bobby Carpenter). The team went 39-25-16 and made the playoffs for the first time, and Langway won the Norris Trophy as the best defenceman.

Teams that made no trades mostly stayed at or under the table’s limit

If the website’s limits are right, the busiest benches should fill right up to them and stop. Washington’s count gives us a way to check. Take every team that, like those Capitals, had no skater play for another NHL team that season. Its count is exact. From 1925-26 to 2020-21, the seasons the table gives a limit for, a team went a whole season like that 156 times.

Some of the busiest land right on the table’s round numbers. The 1973-74 Flyers (50-16-12) and the 1981-82 Jets dressed exactly 17 skaters a night under a 17-skater limit. If 17 was the limit, an average of exactly 17 means every seat was filled in every game. Detroit in 1945-46 sat on exactly 14, and Boston in 1963-64 on exactly 16. In four seasons before 1982-83, and 12 since, the busiest team with no trades used exactly as many skaters as the table allows.

In some seasons the busiest bench filled exactly to the website’s limit; in others it went a little over. One row per season: the busiest team with no trades, in seasons where it sat exactly on the adjustedhockey.com limit (before 1990) or went over it, plus the 1970-71 Flyers, who had trades. * Three skaters on the 1970-71 Flyers also played for another team; this counts only those who played for them alone, so their real figure is at least this high.
SeasonTeamRecordSkater games ÷ team gamesTable limitAgainst it
1936-37New York Rangers19-20-9673 ÷ 48 = 14.02140.02 over
1945-46Detroit Red Wings20-20-10700 ÷ 50 = 14.0014At the limit
1946-47New York Rangers22-32-6899 ÷ 60 = 14.98140.98 over
1947-48Detroit Red Wings30-18-12889 ÷ 60 = 14.82140.82 over

Continues below

Rows 5–14 of 15
SeasonTeamRecordSkater games ÷ team gamesTable limitAgainst it
1948-49Toronto Maple Leafs22-25-13896 ÷ 60 = 14.93140.93 over
1960-61Chicago Blackhawks29-24-171,123 ÷ 70 = 16.04160.04 over
1963-64Boston Bruins18-40-121,120 ÷ 70 = 16.0016At the limit
1964-65Toronto Maple Leafs30-26-141,133 ÷ 70 = 16.19160.19 over
1966-67Toronto Maple Leafs32-27-111,135 ÷ 70 = 16.21160.21 over
1970-71Philadelphia Flyers28-33-171,322 ÷ 78 = 16.95*160.95 over
1973-74Philadelphia Flyers50-16-121,326 ÷ 78 = 17.0017At the limit
1981-82Winnipeg Jets33-33-141,360 ÷ 80 = 17.0017At the limit
1982-83Washington Capitals39-25-161,440 ÷ 80 = 18.0018At the limit
1984-85New Jersey Devils22-48-101,440 ÷ 80 = 18.0018At the limit

Continues below

Row 15 of 15
SeasonTeamRecordSkater games ÷ team gamesTable limitAgainst it
1987-88New Jersey Devils38-36-61,440 ÷ 80 = 18.0018At the limit

Fourteen of the 156 sit above the table’s limit, and they bunch up. Nine come from 1946-47 to 1948-49. The table says 14 for those seasons, but the league as a whole averaged more: 14.6 in 1946-47, 14.6 in 1947-48 and 14.8 in 1948-49. So the table looks wrong for the late 1940s. Four more come from the 1960s, all a fraction over 16: 1960-61 Chicago Blackhawks, 1964-65 Toronto Maple Leafs, 1966-67 Montréal Canadiens and 1966-67 Toronto Maple Leafs (the table shows only the busiest team each season). The odd one out is the 1936-37 New York Rangers, who went over by one player’s appearance in one game across the whole season.

Then there is 1970-71, which sits outside the 156. Three Flyers skaters also played for another team that season, so their count can’t be exact. The skaters who played only for Philadelphia already add up to 16.9 a night, and the traded skaters’ games can only push that higher. The league as a whole averaged 16.6. Both are above the table’s 16, so either that 16 is wrong or the records list skaters in games they never dressed for. The 1970-71 Vancouver Canucks and 1970-71 Pittsburgh Penguins also clear 16 on the same count. The table moves to 17 in 1971-72, so it may be a season late.

From 1949-50 to 1959-60, and from 1971-72 on, no team’s count is above the website’s limit, exact or not. So in most seasons the limits hold up. That still isn’t proof. The exact count works only for teams with no trades, and there were from none to 6 of them a season. And we thought up this test after we had read the table, so we knew what it said before we checked it.

More players per night, many more per season

Washington needed 26 skaters over that season to fill 18 seats a night. Across the century, the number of skaters a team goes through in a season has grown faster than the number it dresses on a night. Count every skater who played for a team at any point in a season (a traded player counts once for each team): from 1917-18 to 1923-24 a team used about 11 skaters in a season; from 2015-16 to 2025-26 it used about 32. The core grew far less. Skaters who played at least 90% of the schedule, nearly every game, went from about 6 a team to about 10.

Teams use about three times as many skaters a season as in the early 1920s; the everyday core grew far less

Every skater who played at least one game, and those who played 90% of the schedule

Show data
Teams used about 11.0 skaters a season in the early 1920s and about 32.0 in the last decade; the number playing 90% of the schedule rose only from about 5.7 to 9.5.
SeasonPlayed at least one gamePlayed 90% of the schedule
1917-1811.3 skaters4.0 skaters
1918-1911.0 skaters4.7 skaters
1919-2011.0 skaters5.5 skaters
1920-2111.0 skaters5.5 skaters
1921-2210.8 skaters6.8 skaters
1922-2310.3 skaters7.0 skaters
1923-2412.0 skaters5.5 skaters
1924-2514.0 skaters7.5 skaters
1925-2614.1 skaters6.9 skaters
1926-2714.8 skaters7.4 skaters
1927-2814.3 skaters7.3 skaters
1928-2914.8 skaters7.5 skaters
1929-3015.1 skaters8.4 skaters
1930-3117.3 skaters8.8 skaters
1931-3217.5 skaters10.4 skaters
1932-3319.0 skaters9.8 skaters
1933-3419.7 skaters9.9 skaters
1934-3520.9 skaters9.3 skaters
1935-3620.8 skaters9.9 skaters
1936-3721.6 skaters9.5 skaters
1937-3820.1 skaters9.9 skaters
1938-3921.1 skaters9.4 skaters
1939-4020.3 skaters7.4 skaters
1940-4120.1 skaters8.0 skaters
1941-4222.0 skaters7.4 skaters
1942-4322.5 skaters8.2 skaters
1943-4424.0 skaters6.5 skaters
1944-4520.7 skaters8.0 skaters
1945-4621.3 skaters8.0 skaters
1946-4723.2 skaters8.8 skaters
1947-4823.5 skaters10.0 skaters
1948-4922.7 skaters10.2 skaters
1949-5025.5 skaters11.5 skaters
1950-5127.2 skaters9.3 skaters
1951-5224.5 skaters8.8 skaters
1952-5324.2 skaters9.7 skaters
1953-5423.7 skaters10.3 skaters
1954-5525.7 skaters11.5 skaters
1955-5623.0 skaters11.0 skaters
1956-5723.0 skaters11.3 skaters
1957-5825.3 skaters11.2 skaters
1958-5921.5 skaters11.5 skaters
1959-6023.3 skaters10.8 skaters
1960-6125.3 skaters11.0 skaters
1961-6222.8 skaters10.3 skaters
1962-6323.2 skaters11.7 skaters
1963-6426.2 skaters10.8 skaters
1964-6526.8 skaters9.7 skaters
1965-6628.7 skaters10.0 skaters
1966-6726.2 skaters9.5 skaters
1967-6825.7 skaters9.3 skaters
1968-6926.0 skaters9.5 skaters
1969-7025.4 skaters9.3 skaters
1970-7127.6 skaters9.2 skaters
1971-7227.3 skaters10.1 skaters
1972-7324.4 skaters10.6 skaters
1973-7427.1 skaters9.3 skaters
1974-7528.0 skaters10.2 skaters
1975-7626.0 skaters10.5 skaters
1976-7726.3 skaters10.7 skaters
1977-7827.8 skaters9.9 skaters
1978-7928.1 skaters8.6 skaters
1979-8030.3 skaters10.0 skaters
1980-8129.6 skaters9.5 skaters
1981-8231.7 skaters7.7 skaters
1982-8330.0 skaters10.6 skaters
1983-8431.9 skaters9.7 skaters
1984-8530.2 skaters9.7 skaters
1985-8632.2 skaters9.6 skaters
1986-8731.9 skaters9.1 skaters
1987-8834.8 skaters8.3 skaters
1988-8934.9 skaters8.2 skaters
1989-9034.0 skaters8.4 skaters
1990-9135.0 skaters8.1 skaters
1991-9236.2 skaters8.3 skaters
1992-9332.5 skaters8.6 skaters
1993-9434.6 skaters8.5 skaters
1994-9531.6 skaters8.7 skaters
1995-9634.8 skaters8.1 skaters
1996-9733.0 skaters8.0 skaters
1997-9833.2 skaters8.2 skaters
1998-9934.5 skaters8.2 skaters
1999-0033.6 skaters8.0 skaters
2000-0132.4 skaters8.5 skaters
2001-0232.7 skaters9.5 skaters
2002-0333.3 skaters9.5 skaters
2003-0434.2 skaters8.3 skaters
2004-05––
2005-0632.3 skaters9.4 skaters
2006-0731.6 skaters10.2 skaters
2007-0830.5 skaters9.3 skaters
2008-0932.0 skaters9.4 skaters
2009-1032.0 skaters9.2 skaters
2010-1132.2 skaters9.6 skaters
2011-1232.3 skaters9.2 skaters
2012-1329.8 skaters9.4 skaters
2013-1431.6 skaters9.0 skaters
2014-1532.3 skaters9.0 skaters
2015-1632.4 skaters9.2 skaters
2016-1731.7 skaters9.5 skaters
2017-1831.1 skaters9.8 skaters
2018-1932.1 skaters9.3 skaters
2019-2030.7 skaters9.5 skaters
2020-2131.4 skaters9.0 skaters
2021-2233.9 skaters8.5 skaters
2022-2332.8 skaters9.7 skaters
2023-2431.1 skaters10.2 skaters
2024-2532.1 skaters10.5 skaters
2025-2632.0 skaters9.2 skaters
  • Second World War: 1939-40 to 1945-46
A traded player counts once for each team he played for. The 90% group uses the player’s season total against the longest team schedule, so it is approximate for traded players. Short seasons (1994-95, 2012-13, 2019-20, 2020-21) have fewer chances to call players up.

On the chart the top line roughly triples, and the bottom one grows far less. The war years pull them apart. In 1943-44 teams used the fewest skaters a game since 1929-30 (about 12) but the most different skaters over a whole season up to then (24 a team), and only about 7 a team played 90% of the schedule, the fewest since 1923-24. The New York Rangers went 6-39-5 and used 32 skaters, and Detroit used 31. Montreal went 38-5-7 with 16, none of whom played for another team. The season totals don’t say why.

The record for one team in one season belongs to the 1991-92 Boston Bruins, who used 51. Fifteen of the players who skated only for Boston that season played 10 games or fewer.

The 60-minute player was gone before anyone timed him

Gardiner’s “every minute” was never timed, and for decades nobody else’s was either. The NHL’s ice-time records begin in 1997-98, so nobody can say from them when the last full-game skater played. They do show he was gone by then. In every season since, the league leader in ice time per game has been a defenceman, averaging between 25:53 and 29:25. That is under half of a 60-minute game. The top mark, 29:25, was Ryan Suter’s in 2013-14. (The fourth-line story covers how the rest of the lineup’s minutes moved.)

Since 1997-98, the ice-time leader has never averaged 30 minutes a game

Minutes per game, skaters who played at least half of the schedule

Show data
The NHL leader in ice time per game has averaged between 25.9 and 29.4 minutes in every season since 1997-98, under half of a 60-minute game.
SeasonLeague leader
1997-9826.6 minutes
1998-9926.7 minutes
1999-0027.7 minutes
2000-0127.8 minutes
2001-0228.4 minutes
2002-0327.5 minutes
2003-0427.5 minutes
2004-05–
2005-0628.3 minutes
2006-0727.6 minutes
2007-0827.3 minutes
2008-0927.0 minutes
2009-1027.4 minutes
2010-1126.9 minutes
2011-1226.9 minutes
2012-1327.3 minutes
2013-1429.4 minutes
2014-1529.1 minutes
2015-1629.0 minutes
2016-1727.4 minutes
2017-1826.8 minutes
2018-1926.7 minutes
2019-2026.0 minutes
2020-2126.4 minutes
2021-2226.2 minutes
2022-2326.4 minutes
2023-2425.9 minutes
2024-2526.8 minutes
2025-2627.7 minutes
  • Reference: Half a game (30.0 minutes)
NHL ice time is published from 1997-98. Every leader since is a defenceman. Overtime minutes are included. The 2004-05 season was cancelled.

The single biggest night came near the end of a lost season. On April 8, 2023, the Anaheim Ducks played at Arizona with five defencemen dressed. Cam Fowler played 38:54. Guinness World Records lists it as the NHL record: “The National Hockey League (NHL) record for most time on ice in a regular season game is 38 minutes 54 seconds.” Matias Maccelli tied the game for Arizona “with 36 seconds left,” ESPN’s recap says, and Barrett Hayton won it “at four minutes of overtime.” It was “the Ducks’ 10th loss in a row.”

Even that night, Fowler was off the ice for about 25 of the game’s 64 minutes. The next-biggest night in our box scores is Thomas Chabot’s 37:50 for Ottawa at Tampa Bay on Dec. 17, 2019, also an overtime loss, 4-3.

Eighteen skaters, nearly every night

The bench, meanwhile, is full nearly every night. From 2017-18 to 2025-26 we have box scores for 22,104 team games (a game between two teams counts once for each of them). A team dressed fewer than 18 skaters in about one of every 316 of them, roughly once every four seasons for a given team. In practice those short nights bunch up: all 70 fell in just 36 team seasons.

Since 2017-18, a team has filled all 18 skater spots in almost every game

From 2017-18 to 2025-26 teams dressed 18 skaters in all but 70 of 22,104 team games; the top five skaters took about 37% of the skater minutes against 27.8% if all 18 played equally.

Team games with fewer than 18 skaters70
of 22,104 (0.32%); the fewest was 15, the most 18
Busiest skater’s minutes24:44–25:04
average per team game, range of season averages
Top five skaters’ share of skater minutes36.7–37.1%
27.8% if minutes were split evenly across 18
Most minutes in one game38:54
Cam Fowler, 2022-23, overtime included; the NHL record since ice time was first tracked in 1997-98 (Guinness)

In those games, a team’s busiest skater played about 25 minutes a night, and that barely changed from season to season. Only 412 of those team games, about 2 in 100, had a skater play 30 minutes. The top of that list, overtime included, is still Cam Fowler’s 38:54 in 2022-23.

A full bench hasn’t spread the work evenly, either. Take all the minutes a team’s skaters played in a game. The top five played about 37 of every 100 of them. If all 18 had split the time evenly, five would have played about 28. Our box scores only go back to 2017-18, and the top five’s share has stayed at about 37 of every 100 minutes in every season since.

No team in any season since 2017-18 came up short more often than the 2020-21 Vegas Golden Knights, who dressed fewer than 18 skaters 10 times. On April 1, 2021, Vegas played Minnesota with 16 and lost in a shootout.

Vegas dressed fifteen

Later that season Vegas went lower still. On May 10, 2021, the Golden Knights played Colorado with ten forwards and five defencemen. “Due to salary cap restrictions and injuries,” Marisa Ingemi wrote for NBC Sports, Vegas was “dressing just 10 forwards and five defensemen.” They had played with 17 for three games. Peyton Krebs had a broken jaw, and because of how much his contract counted against the salary cap (the league’s limit on what a team can spend on players), the team “couldn’t play an additional skater.” As Ingemi put it, “playing with 15 skaters is almost unheard of.” It is the only team game since 2017-18 with fewer than 16.

Shea Theodore played 29:34. At home, with fifteen skaters dressed, the Golden Knights lost 2-1. Fifteen skaters was the whole limit in 1951-52. It was almost twice the eight a night that Ottawa used the year Clancy filled in for everyone.

Share this article

Post on X

Line tools

Live Edgehalla tools on the same subject.

All NHL tools

Show the work

Tests written down
Not in advance (exploratory)
Results computed

Every skater’s games played in every NHL regular season from 1917-18 to 2025-26 (NHL stats API), divided by team games (less the two defaulted 1917-18 Montreal Wanderers games). Rule dates in the rules table are only those we found in a published rule history. Team-level counts use the same season totals: skaters used by one team (a traded player counts for each team), and, for teams that traded no skater during the season, the exact skaters dressed per game (their skaters’ games ÷ team games). Those exact counts are compared with the season limits in adjustedhockey.com’s table (a secondary source with no row-by-row citations; 1925-26 to 2020-21). For teams with traded players, the single-team skaters alone give a lower bound, and a team counts as over the table if even that lower bound is above it. Per-game minutes, dressed counts, defencemen dressed and game length (last play-by-play event) use NHL box scores and play-by-play from 2017-18 to 2025-26. The comparison with the secondary table was done after the memo found it, so it is a check, not a pre-registered test. Exact figures behind the body’s rounded ones: skaters per team game 7.96 in 1917-18; between 17.88 and 18.00 in every season since 1982-83; 10.22 in 1928-29 and 13.04 in 1931-32; 12.23 in 1943-44; 14.58 in 1946-47; 14.69 to 16.08 from 1949-50 to 1969-70; 16.23 to 16.95 from 1970-71 to 1981-82.

Continues below

The 1949-50 step was +0.71 and the 1951-52 step −1.01. Skaters used per team in a season: 11.0 (1917-18 to 1923-24) and 32.0 (2015-16 to 2025-26); playing 90% of the schedule, 5.7 and 9.5. In 1943-44: 24.0 used per team, 6.5 playing 90%. Busiest skater per team game, 2017-18 to 2025-26: 24.73 to 25.06 minutes by season; 412 team games (1.9%) had a 30-minute skater; top five skaters’ share 36.7% to 37.1% against 27.8% if even. How solid the count is. An independent modern check: from 2017-18 to 2025-26 we counted the skaters listed in every team’s box score; the average matches the games-played count to within 0.002 skaters a game in every season. The same series elsewhere: the skaters-per-game figures match the league table used in our NHL GOAT series to four decimal places in every season but 1917-18, where we leave out two defaulted games (forfeits after the Wanderers folded) that the season totals count as played. Both come from the NHL’s season totals, so this checks the arithmetic, not the records. The teams with no trades: for the 156 team-seasons from 1925-26 to 2020-21 in which a team traded no skaters, its skaters’ games divided by its games is an exact count; of those, 14 sit above the secondary table’s limit, all before 1971-72.

Continues below

Early credits are the weak point: before box scores, a games-played credit could mean a player dressed or a player took a shift, and the record keeping of the 1910s and 1920s has gaps (the 1917-18 team totals count two Montreal Wanderers forfeits as games played; we take them out). Read the early line as an approximate count. Short seasons are rates: 1994-95, 2012-13, 2019-20 and 2020-21 were cut short, so their per-game numbers compare, but the season totals of players used do not. What this cannot tell us. The only limit list with every season is an anonymous secondary table, and the rule history we could check lists no skater limit after 1951-52, including the step to 18 in 1982-83. Our no-trade check fits that table from 1949-50 to 1959-60 and from 1971-72 on, but 14 team-seasons and the whole league in 1946-47, 1947-48, 1948-49 and 1970-71 sit above it, so either the table is wrong for those years or the games-played credits count players who never dressed. The exact check covers only no-trade teams, from none to 6 a season. Ice time starts in 1997-98 and our box scores in 2017-18, so Gardiner’s “every minute” is reported, never timed, and Clancy’s night in goal rests on accounts that disagree. Why teams use more players over a season (injuries, call-ups, trades, war) is not in this data, and we did not check why any modern team dressed fewer than 18.

Continues below

Goalies who played per team game: 1.02 in 1964-65 and 1.10 in 1965-66; the body’s “about once in 60” and “about once in 10” are one divided by the amount above one, rounded to the nearest 10. Secondary table limits (adjustedhockey.com) are transcribed by season end year and mapped to start years; 1952-53 has two limits (16 at home, 15 on the road), and the table’s 17 for 1954-55 is its average of 18 through Dec. 1 and 16 after. A team counts as a no-trade team when none of its skaters appears for another NHL team in the season totals; for a team with trades, its single-team skaters alone give a lower bound, and it counts as over the table if even that bound is above it.

Every test we graded
TestResultVerdict
Teams went from about eight skaters a game to 187.96 (1917-18) to 18.00 (2025-26)Holds
Dress-limit changes show up in the count1949-50 +0.71, 1951-52 -1.01; 1953-67 averages exceed the 15 limitPartly

More tests below

Rows 3–5 of 21
TestResultVerdict
Changing on the fly (1927-28) enlarged the bench10.10 to 10.04 skaters a gameNo
Teams with no trades sit at or under the secondary table’s limit14 of 156 team-seasons above it (late 1940s 9, 1960s 4, 1936-37 1); league average above it in 1946-47, 1947-48, 1948-49 and 1970-71; none above it 1949-50 to 1959-60 or from 1971-72; busiest no-trade team exactly at the limit in 4 seasons before 1982-83 and 12 sinceMostly: support, not proof
Busiest no-trade team against the secondary table, 1936-37: New York Rangers (19-20-9)673 ÷ 48 = 14.02 skaters a game; table limit 140.02 over

Continues below

Rows 6–9 of 21
TestResultVerdict
Busiest no-trade team against the secondary table, 1945-46: Detroit Red Wings (20-20-10)700 ÷ 50 = 14.00 skaters a game; table limit 14At the limit
Busiest no-trade team against the secondary table, 1946-47: New York Rangers (22-32-6)899 ÷ 60 = 14.98 skaters a game; table limit 140.98 over
Busiest no-trade team against the secondary table, 1947-48: Detroit Red Wings (30-18-12)889 ÷ 60 = 14.82 skaters a game; table limit 140.82 over
Busiest no-trade team against the secondary table, 1948-49: Toronto Maple Leafs (22-25-13)896 ÷ 60 = 14.93 skaters a game; table limit 140.93 over

Continues below

Rows 10–13 of 21
TestResultVerdict
Busiest no-trade team against the secondary table, 1960-61: Chicago Blackhawks (29-24-17)1,123 ÷ 70 = 16.04 skaters a game; table limit 160.04 over
Busiest no-trade team against the secondary table, 1963-64: Boston Bruins (18-40-12)1,120 ÷ 70 = 16.00 skaters a game; table limit 16At the limit
Busiest no-trade team against the secondary table, 1964-65: Toronto Maple Leafs (30-26-14)1,133 ÷ 70 = 16.19 skaters a game; table limit 160.19 over
Busiest no-trade team against the secondary table, 1966-67: Toronto Maple Leafs (32-27-11)1,135 ÷ 70 = 16.21 skaters a game; table limit 160.21 over

Continues below

Rows 14–17 of 21
TestResultVerdict
Busiest no-trade team against the secondary table, 1970-71: Philadelphia Flyers (28-33-17)1,322 ÷ 78 = 16.95* skaters a game; table limit 16 (* lower bound: traded skaters left out)0.95 over
Busiest no-trade team against the secondary table, 1973-74: Philadelphia Flyers (50-16-12)1,326 ÷ 78 = 17.00 skaters a game; table limit 17At the limit
Busiest no-trade team against the secondary table, 1981-82: Winnipeg Jets (33-33-14)1,360 ÷ 80 = 17.00 skaters a game; table limit 17At the limit
Busiest no-trade team against the secondary table, 1982-83: Washington Capitals (39-25-16)1,440 ÷ 80 = 18.00 skaters a game; table limit 18At the limit

Continues below

Rows 18–21 of 21
TestResultVerdict
Busiest no-trade team against the secondary table, 1984-85: New Jersey Devils (22-48-10)1,440 ÷ 80 = 18.00 skaters a game; table limit 18At the limit
Busiest no-trade team against the secondary table, 1987-88: New Jersey Devils (38-36-6)1,440 ÷ 80 = 18.00 skaters a game; table limit 18At the limit
Box scores agree with games-played credits (2017-26)Within 0.002 skaters a game every seasonHolds
The 60-minute player vanished in the ice-time eraAlready gone: the leader has averaged 25.9-29.4 minutes since 1997-98, always a defenceman; the single-game high is 38:54, and that skater sat 25:06 of a 64:00 gameReframed

Frequently asked questions

How many players dress for an NHL game?

Up to 20: 18 skaters and two goalies. From 2017-18 to 2025-26, teams dressed all 18 skaters in all but 70 of 22,104 team games.

How many players did early NHL teams use?

About eight skaters a game in 1917-18, plus the goalie: five skaters on the ice and about three substitutes. The 1922-23 Ottawa Senators used nine skaters all season and won the Stanley Cup. By the 1950s teams used about 15 a game.

When did NHL teams start dressing 18 skaters?

The count of skaters used per game jumped from 16.95 in 1981-82 to 17.88 in 1982-83 and has been near 18 ever since. One published season-by-season table (adjustedhockey.com, a secondary source) dates the 18-skater limit to 1982-83; the rule history we checked lists no change that season. The 1982-83 Washington Capitals, none of whose skaters played for another team that season, dressed exactly 18.00 skaters a game.

What is the most ice time an NHL skater has played in a game?

Cam Fowler’s 38:54 for Anaheim at Arizona on April 8, 2023, a 5-4 overtime loss. Guinness World Records lists it as the NHL regular-season record since ice time tracking began in 1997-98.

Did King Clancy play every position in one game?

NHL.com says King Clancy played every position during a 1923 Stanley Cup playoff game, including goal while Clint Benedict served a penalty. The historian Eric Zweig, reading the Ottawa Journal, says Clancy played every position over the 1923 playoffs but probably not in one single game; another Journal column said he did, and Zweig calls himself “not quite as sure as I once was.”

What is the fewest skaters an NHL team has dressed in a modern game?

15, by the Vegas Golden Knights against Colorado on May 10, 2021, because of salary cap limits and injuries. It is the only team game from 2017-18 to 2025-26 with fewer than 16 skaters.

More in Minutes

All How Hockey Changed research

Sources and method

  1. NHL stats (NHL.com), every season since 1917-18
  2. NHL stats API: skater and goalie games played, team games, skater time on ice
  3. Scouting the Refs: NHL historical rule changes
  4. adjustedhockey.com: #4. Roster Size (season-by-season limit table; anonymous, secondary)
  5. NHL Rule 5.1 (2013-14 rulebook): no more than 18 skaters and 2 goalies
  6. NHL.com, Stu Hackel: King Clancy, 100 Greatest NHL Players (Jan. 1, 2017)
  7. Eric Zweig: Hockey Nerd in Canada, Part I (Dec. 8, 2014; the Clancy legend and the Ottawa Journal)
  8. Hockey-Reference: King Clancy (born Feb. 25, 1903)
  9. Manitoba Hockey Hall of Fame: Herb Gardiner
  10. Hockey-Reference: Hart Trophy winners (Gardiner, 1926-27)
  11. Russian Machine Never Breaks, Madeline Dement: the Rod Langway trade, 35 years on (Sept. 9, 2017)
  12. Hockey-Reference: Norris Trophy winners (Langway, 1982-83)
  13. Guinness World Records: Most time on ice, NHL regular season game
  14. ESPN recap: Ducks at Coyotes, April 8, 2023
  15. NBC Sports, Marisa Ingemi: Golden Knights will play shorthanded against Avalanche (May 10, 2021)

Research and analysis, not betting advice. Past results do not guarantee future returns; bet only what you can afford to lose.