Edgehalla

How Hockey ChangedThe Clock

The Goalie Pull Keeps Creeping Earlier

Analysts have long argued that coaches should pull the goalie sooner. The data says coaches are listening, a little more each year. We timed every trailing team’s pull from 2017-18 to 2025-26, ranked the coaches, and checked whether the early pullers actually tie more games.

Verdict: Data story

By Edgehalla ResearchPublished 7 min read

More takeaways below

  • Coach rankings are mostly habit, not results. A coach’s pull time repeats from season to season (r = 0.51); his teams’ tie-up rate barely does, and early pullers do not tie measurably more games.
Use the toolNHL empty-net goalsWho scores into and gives up empty nets, season by season.

When do NHL coaches pull the goalie now?

For a team down one goal that had been trailing since at least the five-minute mark, the median pull in 2025-26 came with 2:01 left. In 2017-18 it was 1:41. Down two, the median moved from 2:35 to 3:28, nearly a minute earlier in eight seasons.

Median time left when the goalie is pulled

Regular season; trailing teams that were behind with 5:00 or more left. Bands = middle half of pulls.

Show data
Down one, the median pull moved from 101 to 121 seconds left; down two, from 154.5 to 208 seconds.
SeasonDown twoDown two (range)Down oneDown one (range)
2017-18155 seconds left129 seconds left to 183 seconds left101 seconds left86 seconds left to 122 seconds left
2018-19163 seconds left136 seconds left to 187 seconds left109 seconds left89 seconds left to 130 seconds left
2019-20164 seconds left139 seconds left to 192 seconds left111 seconds left91 seconds left to 128 seconds left
2020-21170 seconds left139 seconds left to 200 seconds left109 seconds left89 seconds left to 125 seconds left
2021-22189 seconds left154 seconds left to 224 seconds left112 seconds left94 seconds left to 136 seconds left
2022-23179 seconds left144 seconds left to 221 seconds left115 seconds left91 seconds left to 137 seconds left
2023-24183 seconds left151 seconds left to 217 seconds left115 seconds left95 seconds left to 137 seconds left
2024-25193 seconds left159 seconds left to 230 seconds left124 seconds left103 seconds left to 145 seconds left
2025-26208 seconds left170 seconds left to 241 seconds left121 seconds left101 seconds left to 142 seconds left
Our goalie-out log from NHL shift data, 2,890 pulls down one and 2,042 down two. Delayed-penalty pulls are excluded.

The trend is steady rather than sudden. Down two, the share of pulls with three or more minutes left rose from 28% in 2017-18 to 70% in 2025-26. Down one, a three-minute pull is still rare (under 6% of pulls in every season), so the change there is a creep from about 1:40 to about 2:00.

More pulls, more empty-net goals

The NHL has published empty-net goals since 2005-06. They went from 0.14 a game then to 0.40 in 2024-25 and 0.39 in 2025-26. Most of the rise came after 2013-14.

Empty-net goals per game

Both teams, NHL regular season

Show data
Empty-net goals rose from 0.14 a game in 2005-06 to 0.40 in 2024-25.
SeasonEmpty-net goals
2005-060.15 goals a game
2006-070.17 goals a game
2007-080.18 goals a game
2008-090.19 goals a game
2009-100.17 goals a game
2010-110.19 goals a game
2011-120.19 goals a game
2012-130.19 goals a game
2013-140.18 goals a game
2014-150.24 goals a game
2015-160.30 goals a game
2016-170.24 goals a game
2017-180.28 goals a game
2018-190.32 goals a game
2019-200.31 goals a game
2020-210.30 goals a game
2021-220.36 goals a game
2022-230.33 goals a game
2023-240.34 goals a game
2024-250.40 goals a game
2025-260.39 goals a game
NHL team totals. Before 2005-06 the game feed marks empty-net goals only by a missing goalie, so a longer series needs the game-by-game backfill.

Does pulling the goalie early pay?

In goals, clearly. Across every one-goal deficit in the last five minutes, trailing teams scored 6.83 goals per 60 minutes with an extra attacker and 2.51 with the goalie in. They also gave up 18.89 empty-net goals per 60. Down one late, a second goal against costs almost nothing, because the game is usually lost either way. That trade is why the math has always favoured pulling sooner.

The trade a pull makes, down one

Goals per 60 minutes for the trailing team, regular season 2017-18 to 2025-26

Trailing by one, teams scored 6.83 per 60 with the goalie pulled and 2.51 with him in, and conceded 18.89 empty-net goals per 60 while pulled.

Goals for (trailing team)
2.5 per 60
6.8 per 60
Empty-net goals against
0 per 60
18.9 per 60
  • Goalie in
  • Goalie pulled
11,626 minutes with the goalie in (last 5:00 only) and 5,339 minutes pulled. Sampling error about ±0.11 and ±0.28 goals per 60.

Yet the league-wide tie-up rate did not move. Of 8,728 one-goal spells that began with five minutes or more left, trailing teams tied it between 39.2% and 41.9% of the time in every season. Twenty extra seconds of six-on-five is worth about one extra goal every 40 pulls, and most deficits never see a pull at all (the trailing team ties it first or time runs out on a late goal). The gain is real but small next to season-to-season noise.

Every coach, ranked

We kept the 32 head coaches with at least 40 pulls down one since 2017-18. Paul Maurice pulls earliest, with a median of 2:11 left; D.J. Smith latest, at 1:32. Read the order loosely: each median rests on 40 to 95 pulls and carries a sampling error of roughly ±4 to ±8 seconds, and the league itself moved 20 seconds in these years, so coaches whose tenure is recent look earlier than coaches of the same habit from 2017-18. Measured against each season’s league median, Maurice is still the earliest (about 17 seconds ahead of the norm) and Don Granato and D.J. Smith the latest, while Andrew Brunette, second on raw time, drops to the middle of the top group. Most of the middle of the table is within noise of its neighbours. The scatter asks whether early pullers tie more games.

Early pullers do not clearly tie more games

Head coaches with 40+ pulls down one, regular season 2017-18 to 2025-26

Show data
Across 32 coaches, median pull time and the share of one-goal deficits tied have a correlation of 0.19 (95% range -0.17 to 0.50).
Median pull, seconds leftOne-goal deficits tied
Paul Maurice131 s41%
Andrew Brunette130 s41%
Jon Cooper129 s42%
Rod Brind'Amour127 s45%
Lindy Ruff127 s39%
John Tortorella126 s38%
Peter Laviolette125 s50%
Dean Evason125 s43%
Bruce Boudreau124 s41%
Sheldon Keefe121 s44%
David Quinn121 s35%
Joel Quenneville119 s44%
Gerard Gallant118 s47%
Jared Bednar117 s42%
Jim Montgomery116 s47%
Peter DeBoer115 s45%
André Tourigny115 s38%
Todd McLellan112 s36%
Rick Tocchet111 s38%
Craig Berube110 s41%
Jeff Blashill110 s27%
Rick Bowness107 s43%
Barry Trotz107 s39%
Travis Green106 s37%
Bruce Cassidy105 s51%
Dave Hakstol103 s38%
Mike Sullivan103 s48%
John Hynes103 s38%
Martin St. Louis102 s45%
Bob Boughner101 s36%
Don Granato94 s42%
D.J. Smith92 s35%
Tie-up rate = share of one-goal deficits (trailing with 5:00+ left) the team tied, pulled or not. Each coach’s rate has a sampling error of about ±3 to ±5 points, and it also reflects how good his teams were.
Head coaches with 40+ pulls down one (regular season, 2017-18 to 2025-26), earliest first.
CoachSeasonsDeficitsPullsMedian pullDeficits tied
Paul Maurice9217822:1141.0%
Andrew Brunette4134432:1041.0%
Jon Cooper9262882:0942.4%
Rod Brind'Amour8217682:0744.7%
Lindy Ruff6199592:0738.7%
John Tortorella7207622:0638.2%
Peter Laviolette8219752:0550.2%
Dean Evason7150412:0543.3%
Bruce Boudreau5125412:0440.8%
Sheldon Keefe7186512:0144.1%
David Quinn5148482:0134.5%
Joel Quenneville6145461:5944.1%
Gerard Gallant5136501:5847.1%
Jared Bednar9237741:5742.2%
Jim Montgomery6151471:5647.0%
Peter DeBoer9227731:5544.9%

Continues below

Rows 17–32 of 32
CoachSeasonsDeficitsPullsMedian pullDeficits tied
André Tourigny5197641:5537.6%
Todd McLellan9222831:5236.0%
Rick Tocchet8231901:5138.1%
Craig Berube8212641:5041.0%
Jeff Blashill6199711:5027.1%
Rick Bowness6136431:4742.6%
Barry Trotz5127461:4739.4%
Travis Green8199691:4636.7%
Bruce Cassidy9249721:4551.4%
Dave Hakstol5135501:4337.8%
Mike Sullivan9278951:4347.8%
John Hynes9250761:4338.4%
Martin St. Louis5164451:4245.1%
Bob Boughner5153451:4135.9%
Don Granato4108401:3441.7%
D.J. Smith6130511:3234.6%

What is signal and what is noise

To check, we split each coach’s deficits into alternate seasons (2017-18, 2019-20 and so on against the rest). Pull time repeats: a coach who pulls early in one half pulls early in the other (r = 0.51, 95% range 0.16 to 0.74, 27 coaches). Measuring each pull against that season’s league median, so the league-wide drift cannot do the work, makes the habit link stronger, not weaker (r = 0.62). Tie-up rate barely repeats (r = 0.26, range -0.14 to 0.58).

Show the work

Tests written down
Not in advance (exploratory)
Results computed

Every third-period trailing spell in the regular season, 2017-18 to 2025-26, from our goalie-out log built on NHL shift data (8,728 one-goal spells that began with 5:00+ left). Coach tests use the 32 head coaches with 40+ pulls, with alternate seasons as the holdout.

Every test we graded
TestResultVerdict
Pulls are getting earlierDown one 101 s to 121 s; down two 154.5 s to 208 sHolds
A coach’s pull time repeatsr = 0.51 (0.16 to 0.74)Holds
Early-pulling coaches tie more deficitsr = 0.19 (-0.17 to 0.50)Noise
A coach’s tie-up rate repeatsr = 0.26 (-0.14 to 0.58)Weak, not significant

Frequently asked questions

When do NHL teams pull the goalie?

In 2025-26 the median pull for a team down one came with about two minutes left (2:01); down two, about 3:28. Both are earlier than in 2017-18, when they were 1:41 and 2:35.

Does pulling the goalie early work?

It adds goals: down one late, teams score about 6.8 goals per 60 minutes with an extra attacker against 2.5 with the goalie in. But the effect on tying games is small, and across nine seasons the league-wide tie-up rate has not visibly moved.

Which NHL coach pulls the goalie earliest?

Among head coaches with at least 40 pulls down one since 2017-18, Paul Maurice has the earliest median pull, at about 2:11 left. D.J. Smith has the latest, at 1:32. Neighbouring coaches in the ranking are within a few seconds of each other, inside sampling error, but against each season’s league norm Maurice is still the earliest and Smith and Don Granato the latest.

How many empty-net goals are there per game?

About 0.4 for both teams combined: 0.40 in 2024-25 and 0.39 in 2025-26, up from 0.14 in 2005-06.

Sources and method

  1. NHL stats (NHL.com), every season since 1917-18
  2. NHL stats API: team empty-net goals (team realtime totals), 2005-06 on
  3. Hockey-Graphs: the state of goalie pulling in the NHL (2020)

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