How Hockey ChangedMinutes
Stuck After an Icing, NHL Defences Give Up 15% More Expected Goals
When a team can’t get its tired players off the ice, does it pay in goals? Yes, a little. For each of the 747,958 5-on-5 shots since 2017-18 that got past the shot-blockers, we checked how long the defenders had been on the ice. Shots against a defence late in its shift are more dangerous, but about half of that is because long shifts happen when a team is pinned in its own end. The icing rule keeps the same five skaters on the ice, so it is the fairest test of tired legs on their own, and it shows a small but real cost.
By Edgehalla ResearchPublished 14 min read
At 17:23 of overtime in Game 6, Pittsburgh iced the puck, and the rules kept the same five Penguins on the ice. When the puck was dropped, Ben Kindel, Erik Karlsson and three teammates had already been out for 87 seconds.
Kindel lost the draw to Noah Cates. Nine seconds later Cam York’s shot from the point, out by the blue line, went in, and Philadelphia had won the series.
“We were a shot away from going back to Pittsburgh for Game 7,” Sidney Crosby said. “It comes down to bounces sometimes.”
On this page
Long shifts end with more dangerous shots
That was April 29, 2026, in Philadelphia. The Penguins had lost the first three games of the series and won the next two. Counting the nine seconds after the draw, the five Penguins had been out for 96 seconds when York’s shot went in. In nine seasons of playoff overtime, no five-on-five goal had come against a defence that had been out longer.
Crosby called it a bounce. This story asks the bigger question behind his answer: when a team can’t get its tired players off the ice, does it pay in goals? It does, a little. Shots against a tired defence are more dangerous. About half of that comes from where long shifts happen. The cleanest test of tired legs on their own, the icing rule, puts their cost at about two thirds of a goal per team a season.
To get there we timed every 5-on-5 shot from 2017-18 to 2025-26 that got past the shot-blockers, whether it was on goal, wide or off the post. For each shot we checked how long each of the five defending skaters had been on the ice, and took the average. The 747,958 shots produced 44,540 goals.
To rate each shot we use expected goals: how often a shot like it goes in, from where and how it was taken. Philippe Myers’s overtime winner for Philadelphia in 2020 rated about a one-in-70 chance. Alex Ovechkin’s overtime winner against Montreal in 2025 rated about one in three. A shot that goes in about one time in 20 counts as 0.05 of a goal. Add up a defence’s shots against and you get its “danger”, the goals an average finisher would score from them. Nothing in the rating knows how long the defenders had been out.
Against a defence that had just come on, about 4 shots in 100 would be expected to go in. Against a defence 80 seconds or more into its shift, about 7 would. Real goals followed the same path. About a third of all 5-on-5 goals came against a defence 40 seconds or more into its shift, and about one in 14 against a defence a minute or more in.
The longer a defence has been out, the more dangerous the shots against it
Goals you would expect from 100 5-on-5 shots that weren’t blocked, and goals actually scored, by how long the defenders had been on the ice, 2017-18 to 2025-26
per 100 shots
Defenders’ average seconds into the shift
Show data
| Defenders’ average seconds into the shift | Expected goals | Actual goals |
|---|---|---|
| 5 | 4.2 per 100 shots | 3.8 per 100 shots |
| 15 | 5.3 per 100 shots | 5.1 per 100 shots |
| 25 | 5.9 per 100 shots | 5.9 per 100 shots |
| 35 | 6.3 per 100 shots | 6.2 per 100 shots |
| 45 | 6.5 per 100 shots | 6.4 per 100 shots |
| 55 | 6.7 per 100 shots | 6.7 per 100 shots |
| 65 | 6.7 per 100 shots | 6.7 per 100 shots |
| 75 | 6.9 per 100 shots | 7.0 per 100 shots |
| 85 | 7.0 per 100 shots | 7.1 per 100 shots |
The curve makes tired legs look worse than they are, because its two ends are different kinds of hockey. A defence that has just come on mostly faces shots soon after a faceoff in its own end, and those rarely go in. A defence that has been out a minute is usually deep in a long scramble in its own end. The table shows how the shots change along the way.
| Defenders’ time on the ice | Shots | Within 5 s of a faceoff in the defence’s own end | Play began after the defence iced it | Rush | Rebound |
|---|---|---|---|---|---|
| 0-10 s | 41,408 | 29.5% | 3.4% | 13.7% | 5.2% |
| 20-30 s | 179,256 | 2.2% | 8.2% | 9.7% | 7.6% |
| 40-50 s | 120,287 | 1.3% | 12.0% | 8.2% | 9.1% |
| 60-70 s | 26,683 | 1.4% | 17.9% | 4.6% | 10.3% |
| 80 s+ | 8,695 | 1.5% | 28.7% | 2.7% | 10.3% |
Read down the columns. Nearly a third of the shots against a fresh defence came within five seconds of a draw in its own end. Past 40 seconds almost none did. Rebounds about doubled, from about 5 shots in 100 to about 10. Rush shots went the other way, from about one in 7 to about one in 37. The after-icing column climbs fastest: about 3 shots in 100 against a fresh defence, nearly three in ten past 80 seconds. Many of the longest shifts are icing shifts.
About half of that is where long shifts happen
So the first job is to compare like with like. We put every shot in a group with similar shots: same kind of faceoff before it, same time since that faceoff, whether the play followed an icing, and the same score. Then we asked whether, inside each group, shots against a tired defence were still more dangerous than shots against a fresh one.
They were, but the gap shrinks. Before the matching, shots against a defence 80 seconds in were about 68% more dangerous than shots against one that had just come on. Matching on the setting brings that down to about 31%, so the setting alone accounts for about half of the raw gap. If we also match on whether the attacking skaters were fresh themselves, and on whether the shot was a rebound or came off a rush, the gap falls to about 23%, about a third of where it started.
Every season showed it. Take every defence 40 seconds or more into its shift and compare it with fresher ones facing similar shots. In each of the nine seasons, the longer-shift defence gave up more danger, by between about 6% and 13% depending on the year.
Even against similar shots, a defence past 80 seconds gives up about 31% more danger than a fresh one
Danger of each shot compared with similar shots from the same kind of moment: 100 = average for that kind of moment; 110 = 10% more dangerous
Defenders’ average seconds into the shift
Show data
| Defenders’ average seconds into the shift | Same setting | Same setting (range) | Closer match (attackers’ rest, rebounds, rushes) |
|---|---|---|---|
| 5 | 86 | 85 to 87 | 90 |
| 15 | 92 | 92 to 93 | 94 |
| 25 | 98 | 97 to 98 | 98 |
| 35 | 102 | 101 to 102 | 101 |
| 45 | 105 | 104 to 105 | 103 |
| 55 | 107 | 106 to 108 | 106 |
| 65 | 108 | 107 to 110 | 106 |
| 75 | 110 | 108 to 113 | 108 |
| 85 | 113 | 111 to 116 | 110 |
- Reference: Average for the setting (100)
Read the chart from left to right. Shots against a defence that had just come on were about 14% less dangerous than the average shot from the same kind of moment. At 40 to 50 seconds they were about 5% more dangerous, and past 80 seconds about 13% more. Now put the two ends side by side, 113 against 86 on the chart. The late-shift shot is about 31% more dangerous than the fresh one, the gap after matching from above.
Those shots go in about as often as their danger says
Tired legs could cost goals one more way. If tired defenders also let good shots through more often, more shots against them would go in than their danger rating says. Over the nine seasons they didn’t. Against defences 40 to 50 seconds in, about 100 goals went in for every 100 the shots were worth. Past 80 seconds it was about 100. So tired defences pay through the shots they allow, such as the extra rebounds in the table above. Once a shot is taken, it goes in at about its normal rate.
A pinned defence also faces shots faster
About a third of all 5-on-5 goals came against defences 40 seconds or more into a shift, yet those defences are on the ice for less than a fifth of the time. Mostly, they are under siege. We rebuilt every second of 5-on-5 play over the nine seasons from the league’s on-ice records, about 537,000 minutes, each counted twice, once with each team defending. For every second we knew how long the defence had been out.
A defence past 80 seconds faces about five times as many shots a minute as a fresh one
Shot attempts against that got past the shot-blockers, per 60 minutes of 5-on-5 play at each point in the shift, 2017-18 to 2025-26
attempts per 60
Defenders’ average seconds into the shift
Show data
| Defenders’ average seconds into the shift | Attempts against per 60 |
|---|---|
| 5 | 21 attempts per 60 |
| 15 | 30 attempts per 60 |
| 25 | 38 attempts per 60 |
| 35 | 47 attempts per 60 |
| 45 | 61 attempts per 60 |
| 55 | 73 attempts per 60 |
| 65 | 86 attempts per 60 |
| 75 | 97 attempts per 60 |
| 85 | 106 attempts per 60 |
A defence in its first 10 seconds faced about 21 attempts per 60 minutes of play, the length of a full game. At 40 to 50 seconds it faced about 61, and past 80 seconds about 106, roughly five times the fresh rate. Most of that is about where the puck is. A team under attack in its own end can’t get to the bench, so the longest shifts are the ones spent defending. We don’t turn this curve into a fatigue number. It does show where the danger piles up. On the ice for less than a fifth of 5-on-5 play, defences 40 seconds or more into a shift face about three in ten shot attempts and about a third of the danger.
The NHL wrote a rule to keep tired players on the ice
Hockey has one moment when the rules, not the coach, decide who stays on. Since 2005-06, a team that ices the puck may not change before the faceoff that follows. The rulebook says that team “shall not be permitted to make any player substitutions prior to the ensuing face-off.” The other team may change, and it chooses which of the two faceoff dots in the icing team’s end the puck is dropped on.
The idea, as ESPN’s Pierre LeBrun put it in 2017, “was to try and take advantage of that tired team and create offense as a result.” Coaches found a way out. They called their one timeout at the icing and gave the stuck players a rest.
On March 8, 2017, at their meeting in Boca Raton, Fla., the general managers voted to recommend taking that away. “Over time it morphed into using it as a way to rest,” said Kevin Cheveldayoff, the Winnipeg Jets’ general manager. Gary Bettman, the commissioner, said that if you ice the puck “there should be a consequence to it,” and named the consequence: “having players who aren’t rested.” The league’s list of rule changes for 2017-18 put it in one line: “No time-out shall be granted to the team that commits an icing infraction.”
The ban took hold. We checked the official reports for a sample of 24 games from each season, before and after it. Before 2017-18, a team called a timeout right after an icing about once every four games. Since then it has happened about once every 36 games, and those timeouts belong to the attacking team, which the rule still allows.
Our shift data starts in 2017-18, the first season of the ban. So every stuck defence in this story had to stay out without a timeout of its own.
A stuck unit can still catch its breath now and then. The other team may call its own timeout, and a TV timeout sometimes falls at an icing. In the sampled games since 2017-18, six of 1,864 icing stoppages came with a team timeout, each called by the attacking team, and one came with a TV timeout. Brayden Point’s goal, later in this story, came right after a TV break at an icing.
The longest-stuck defence to give up a 5-on-5 regular-season goal in that span was in Montreal, on Dec. 4, 2023. Brendan Gallagher, Tanner Pearson, Christian Dvorak, Mike Matheson and Gustav Lindstrom went over the boards at 11:09 of the second period. The NHL’s official game log shows Montreal icing the puck at 13:24, so all five had to stay for the faceoff in their own end. Seattle sent out a different five. Matheson lost his stick, as the Field Level Media recap put it, and Jared McCann scored 41 seconds after the draw. The Canadiens had been out for 2:56, nearly three minutes. They still won the game 4-2. One goal proves nothing on its own. Across every stuck faceoff since the ban, did Bettman get his consequence?
Stuck after an icing, a defence pays more the longer it was out
He did. We compared 86,919 5-on-5 faceoffs in a team’s own end right after its own icing with 232,937 other faceoffs in its own end, and counted what the attackers created in the next 20 seconds. After an icing, the attackers created about 15% more danger: 1.66 expected goals per 100 draws, against 1.44 after other faceoffs in that end. The stuck team also lost more draws: it won about 47 in 100, against about 52 normally.
After an icing, the longer the stuck players had been out, the more danger they allowed
Expected goals the attackers created in the 20 seconds after a 5-on-5 faceoff in the defending team’s end, per 100 faceoffs, 2017-18 to 2025-26
After its own icing, a defence whose skaters had been out under 20 seconds allowed 1.49 expected goals per 100 draws, close to the 1.44 after other defensive-zone draws. Past 60 seconds it allowed 1.99.
The longer the stuck players had been out, the worse it got. When they had been out a minute or more, the attackers created about 38% more danger than after an ordinary draw in that end. The stuck centre lost more faceoffs too: his team won about 49 draws in 100 when the players had been out under 20 seconds, and about 44 past a minute.
Over nine seasons, the extra danger after icings comes to about 190 goals, if the shots went in at their usual rates. That is about 21 a season across the league, or about two thirds of a goal per team a season. It takes about 450 stuck faceoffs to add up to one extra goal at usual finishing rates. Count the goals that actually went in, and stuck defences gave up about 270 more than they would have after ordinary draws, about 22% more: 1.70 goals per 100 stuck draws, against 1.39. That is more than the shot danger predicted, but the difference could be luck.
Real goals are too few to show that climb by themselves. The minute-plus icing group produced just 126 goals in the 20 seconds after the draw. Per 100 stuck draws, real goals ran about 1.6 when the stuck players had been out under 20 seconds, 1.7 at 20 to 40 seconds, 1.7 at 40 to 60 and 2.1 past a minute. Flat in the middle, then up. That is too jumpy to read alone, so the danger figures carry the pattern. The fairest comparison for a long stuck shift would be a long shift with no icing, and that group is thinner still: only 1,486 faceoffs in a team’s own end came with the defence 40 seconds or more into its shift and no icing.
The cost shows up every season, and it isn’t the icing itself
We checked each season on its own, in case one or two odd years drove the result. The faceoff after a team’s own icing was more dangerous than other faceoffs in its end in all nine seasons, by between about 10% and 23%.
The faceoff after an icing was more dangerous in every season
Expected goals the attackers created in the 20 seconds after a 5-on-5 faceoff in the defending team’s end, per 100 faceoffs
expected goals per 100 draws
Show data
| Season | After own icing | Other draws |
|---|---|---|
| 2017-18 | 1.31 expected goals per 100 draws | 1.20 expected goals per 100 draws |
| 2018-19 | 1.42 expected goals per 100 draws | 1.27 expected goals per 100 draws |
| 2019-20 | 1.79 expected goals per 100 draws | 1.52 expected goals per 100 draws |
| 2020-21 | 1.74 expected goals per 100 draws | 1.47 expected goals per 100 draws |
| 2021-22 | 1.86 expected goals per 100 draws | 1.59 expected goals per 100 draws |
| 2022-23 | 1.81 expected goals per 100 draws | 1.61 expected goals per 100 draws |
| 2023-24 | 1.74 expected goals per 100 draws | 1.51 expected goals per 100 draws |
| 2024-25 | 1.67 expected goals per 100 draws | 1.35 expected goals per 100 draws |
| 2025-26 | 1.70 expected goals per 100 draws | 1.47 expected goals per 100 draws |
When the stuck players were fresh, the icing barely hurt them. With the players out under 20 seconds, the faceoff after an icing was less than 5% more dangerous than normal. The cost climbs with the stuck skaters’ time on the ice.
That climb held when we split the nine seasons in two. From 2017-18 to 2021-22, a stuck defence gave up about 1.4 expected goals per 100 draws when its skaters had been out under 20 seconds, and about 1.9 past a minute. From 2022-23 to 2025-26 the same two figures were about 1.6 and 2.1. Ordinary draws in a team’s own end ran about 1.4 and 1.5 in the two halves.
Now compare two defences that have both been out 20 to 40 seconds. One stayed by choice. After any other whistle in its own end, such as a goalie freezing the puck, a team is free to change, and sometimes the coach sends the same skaters back out. We found 19,488 of those faceoffs. The other defence was trapped by an icing. The trapped one gave up about 8% more danger.
One thing the data can’t split is tiredness from the matchup, meaning which players end up facing which. After an icing, the attacking coach chooses who goes out. The players who had been out longest may also be the ones who got pinned in the first place.
The playoffs charge the same price
The playoffs tell the same story on a smaller sample. In 7,503 playoff faceoffs after a team’s own icing, the attackers created about 15% more danger than after other faceoffs in that end, the same gap as in the regular season. The stuck team won about 47 of those draws in 100, against about 53 normally. The danger rose with the legs here too: about 1.5 expected goals per 100 draws when the stuck skaters had been out under 20 seconds, 1.7 at 20 to 40 seconds, 1.8 past 40. Those steps point the same way as the regular season, but on playoff numbers this small they could be luck.
Real playoff goals after an icing ran much higher, about 56% above ordinary draws. The stuck defences themselves gave up 134 goals on shots worth about 124, only a little more than their danger said. The big gap comes from the other side of the comparison. Ordinary playoff faceoffs in a team’s own end happened to run low: 202 goals on shots worth about 254, about four goals for every five their shots deserved. So we treat the bigger goal gap as luck.
Overtime alone is thinner still. In 469 overtime faceoffs after the defending team’s own icing, the attackers created about 18% more danger than after other overtime draws in that end. That is too few faceoffs to lean on by itself.
In overtime, one stuck shift can end a series
Some of those stuck shifts ended games. On Aug. 11, 2020, in an empty arena in the Toronto bubble, Tampa Bay and Columbus reached the fifth overtime of Game 1. The Blue Jackets “already had tired legs,” ESPN’s Emily Kaplan wrote, “playing their sixth game in nine days.” Columbus had just killed a penalty when, at 10:11 of the fifth overtime, it iced the puck. The NHL’s official game log records the stoppage as “ICING,TV TIMEOUT”. The stuck Blue Jackets got a television break, but no change. Brayden Point beat Riley Nash on the draw and scored 16 seconds later, 150 minutes and 27 seconds into the game. By then David Savard and Vladislav Gavrikov had been out for 92 seconds, and the five Columbus skaters averaged 70 seconds.
The Lightning were spent as well. Tampa Bay’s Yanni Gourde said afterward:
“We were all exhausted. We were all looking for a goal.”
Point said there was no plan to it: “I see a rolling puck coming to me, I just throw it on net. I’m not even thinking.” Joonas Korpisalo had made 85 saves, a playoff record, and Seth Jones had played 65 minutes and 6 seconds, another.
Philadelphia did it twice in the same bubble, beating the Islanders in overtime right after an Islanders icing: Philippe Myers on Aug. 26 and Scott Laughton on Sept. 1, against defences about 67 and 85 seconds into their shifts. Among all 148 five-on-five playoff overtime goals in our nine seasons, Laughton’s came against the second-longest-serving defence. Only York’s Penguins had been out longer. On April 21, 2025, Alex Ovechkin won Game 1 against Montreal in overtime six seconds after a Montreal icing, against a defence 62 seconds in. Counting every playoff overtime goal from 2017-18 to 2025-26, power plays included, 177 in all, eight came within 20 seconds of a faceoff after the losing team’s own icing. Five of those eight came against a defence a minute or more into its shift, York’s among them.
Teams haven’t changed their habits to dodge the cost
A price of two thirds of a goal a season is small enough to live with, and so far teams have. Coaches could shorten shifts so fewer shots come against tired legs. In this window they haven’t. Every season since 2017-18, about three in ten 5-on-5 shot attempts came against a defence 40 seconds or more into its shift. The average shot came with the defenders about 33 seconds in, every season.
Every season, tired defences played under a fifth of the time but faced about three in ten shots
Defences 40 seconds or more into a shift: share of 5-on-5 time, shot attempts and expected goals against
Show data
| Season | Expected goals against | Attempts against | Time on ice |
|---|---|---|---|
| 2017-18 | 32% | 29% | 19% |
| 2018-19 | 33% | 30% | 19% |
| 2019-20 | 34% | 31% | 19% |
| 2020-21 | 33% | 31% | 19% |
| 2021-22 | 33% | 31% | 18% |
| 2022-23 | 35% | 32% | 18% |
| 2023-24 | 33% | 30% | 18% |
| 2024-25 | 34% | 30% | 18% |
| 2025-26 | 35% | 31% | 19% |
The time and shot lines run flat. The danger line edges up by a point or two. Nine seasons is a short window, though, and it starts long after the era of two-minute shifts. Whether shifts got shorter over the decades would need shift charts from before 2017-18, which this story does not use.
No player has a knack for tired legs
If the cost is steady, maybe some players cash in on it more than others. None stands out. Auston Matthews has scored the most 5-on-5 goals against a defence 40 seconds or more into its shift in these nine seasons, 92 of his 241. That is about 38% of his goals, against about 34% for the league, a gap small enough to be luck. Among the 109 players with 100 goals at 5-on-5, the share runs from Anthony Beauvillier (about 43%) to Nino Niederreiter (about 23%). That is about the spread luck alone would give that many players.
The best test is whether a habit lasts. We split the nine seasons in two, 2017-18 to 2021-22 and 2022-23 to 2025-26. A scorer’s share of goals against long-shift defences in the first half told us nothing about his share in the second. His share of goals scored right after an opponent’s icing told us almost nothing either. Jake Guentzel scored the most of those, 15. Across the league about one 5-on-5 goal in 30 comes that way, which would have given him about six. Most of his came early: 11 among his first 95 goals at 5-on-5, more than three times the league rate. Then he scored four among his next 75, much closer to it. A real knack would have lasted. His didn’t.
Centres tell the same story from the other side. Claude Giroux won about 63% of his 542 faceoffs after his own team’s icing, the best of the 69 centres who took at least 400 of them. He also won about 63% of his other faceoffs in his own end. Good faceoff men stay good from one half of the data to the other. A centre’s edge on stuck draws over his other draws in one half told us little about the next.
Tired legs do cost goals, just not many
Put the pieces together and the answer to the question is yes, a little.
- Tired legs cost something you can measure. The cleanest test, the icing rule, puts it at about 15% more danger in the 20 seconds after the faceoff, rising to about 38% when the stuck skaters had been out a minute or more.
- Matching each shot with similar ones takes away about half of the extra danger late in a shift. That half comes from the situation: long shifts happen when a team is pinned in its own end. Part of what is left may still be that.
- Teams kept the same habits for nine seasons. How long the defenders had been out when a shot came, and how much of the game is played on tired legs, have stayed flat since 2017-18.
- Shots against tired defences go in about as often as their danger says. A recent break, from 2023-24 on, lines up with a change in how the league records shots, so it stays open.
- The playoffs charge the same price, about the same extra danger after an icing, on a smaller sample.
- No scorer cashes in on tired defences more than luck would explain, and no centre has a special touch on the stuck faceoff.
Two limits stay with all of it. Tired and pinned can’t be fully pulled apart, and the single games in this story prove nothing on their own. How each number was built is under “Show the work”.
Five Penguins, 96 seconds
Bettman got his consequence. A stuck defence gives up more, about two thirds of a goal per team a season, and more the longer it has been out. Crosby had a case too. York shot from the point, and by our measure the shot had about a one-in-30 chance of going in.
On a single night the cost of tired legs looks like a bounce. Pittsburgh’s came due on April 29 in Philadelphia, on five Penguins who had been on the ice for 96 seconds.
Share this article
Power play tools
Live Edgehalla tools on the same subject.
- Power play tonight PP1 units and power-play matchups for tonight’s games.
- NHL line combinations today Forward lines, defence pairs and PP units for all 32 teams.
- NHL referees Tonight’s officials and how many penalties each crew calls.
- Fantasy hockey projections Tonight’s projected stats for every skater and goalie.
Show the work
- Tests written down
- Not in advance (exploratory)
- Results computed
Every 5-on-5 regular-season unblocked shot attempt from 2017-18 to 2025-26 (747,958), with the defending skaters’ average shift age from NHL shift charts. A skater’s shift age is game-clock seconds since his continuous shift began (shifts with a gap of a second or less are chained); the defence’s age is the mean over its five skaters at the event. Expected goals come from our shot model (location, shot type, previous event, rebounds and rush, score, time), which has no shift-age input. Adjusted comparisons use indirect standardisation within settings: the zone of the faceoff that began the play, seconds since that faceoff (0-5, 5-15, 15-30, 30-60, 60+), whether the play followed an icing, and the score (from down two to up two). Within each setting the average shot’s expected goals, pooled over all shift ages, is the reference; each shift-age bucket’s index is its observed expected goals ÷ the reference expected goals of its shots. The strict version also matches the attackers’ shift age and the rebound and rush flags. Exposure is rebuilt second by second from on-ice stints (a player’s shift = consecutive stints he is on for; ages grow one second per second of play), each second counted twice, once with each team defending. Holdouts: every season separately, and 2017-18 to 2021-22 against 2022-23 to 2025-26.
Continues below
Excluded as a feed artefact: in Buffalo-New Jersey on March 29, 2024, the NHL shift chart has all ten skaters on from about 4:00 to 7:48 of the first period without a change, which would put one shot attempt against a defence 228 seconds into its shift. We removed that stretch (1 unblocked attempt, no goals, 412 defender-seconds of exposure); other attempts past three minutes were not checked against the NHL’s reports. Icing test: 5-on-5 faceoffs in the defending team’s own end, split by whether the stoppage before the draw was that team’s own icing, with the attackers’ expected goals and goals in the 20 seconds after the draw. Extra totals compare the stuck draws with the other-draw rate: extra goals from exact counts, extra expected goals from the two-decimal rates (so shown to the nearest 10); per team uses the mean of 31.56 teams a season. Playoffs: the same test rerun on 2017-18 to 2025-26 playoff games. Players: split-half persistence of each scorer’s share of 5-on-5 goals against defences 40 seconds or more in and right after an opponent’s icing (scorers with 50+ goals in each half), and of each defending centre’s win rate on stuck draws (centres with 150+ stuck draws in each half); single-player shares are compared with the league share by a z-score. Scenes come from our play-by-play and the NHL’s HTML reports; they illustrate and are not tested.
Continues below
Timeouts at icings come from the official reports of the 24 games a season sampled for our stoppages story.
| Test | Result | Verdict |
|---|---|---|
| Shots against stale defences are more dangerous, like for like | +6.1% to +12.6% in each of 9 seasons; index 86.1 (under 10 s) to 113.1 (80 s+); strict 89.7 to 110.1 | Holds, partly selection |
| Raw shot danger rises with defence shift age | 4.2 (under 10 s), 6.5 (40-50 s), 7.0 (80 s+) expected goals per 100 shots; 33.6% of 5v5 goals vs 40 s+ defences, 7.2% vs 60 s+ | Descriptive (setting confound) |
| Being stuck after an icing costs expected goals | 1.66 vs 1.44 xG per 100 draws in 20 s (n 86,919 / 232,937); ahead in 9 of 9 seasons (9.5% to 23.3%); halves 1.61 / 1.73 vs 1.40 / 1.49; about 190 extra xG, 21.2 a season, 0.67 per team-season (mean 31.56 teams), one per ~450 stuck draws | Holds |
More tests below
| Test | Result | Verdict |
|---|---|---|
| Actual goals after an icing beat the expected-goals gap | 1.70 vs 1.39 goals per 100 draws; about 270 goals above the other-draw rate; the part above expected goals is within noise | Not shown |
| Stuck teams lose more draws | 47.3% vs 52.2%; 48.9% (under 20 s) to 43.5% (60 s+) | Holds |
| The icing cost grows with the stuck skaters’ shift age | 1.49 (under 20 s), 1.63, 1.82, 1.99 (60 s+); 2017-22 1.44 to 1.87, 2022-26 1.57 to 2.11; actual goals 1.63, 1.68, 1.68, 2.11 (too noisy alone) | Holds |
| Forced stays cost more than voluntary ones at the same shift age | 20-40 s: 1.63 after icing vs 1.51 without (n 19,488) | Holds (matchup not separable) |
Continues below
| Test | Result | Verdict |
|---|---|---|
| The icing cost holds in the playoffs | 1.65 vs 1.44 xG per 100 draws (n 7,503 / 17,608); win 47.1% vs 52.8%; by stuck age 1.53 / 1.66 / 1.76; overtime 1.94 vs 1.65 (n 469); goals 1.79 vs 1.15, but other draws scored 0.80 goals per expected goal | Holds (smaller sample); goal gap is noise |
| Playoff overtime goals right after the loser’s own icing | 8 of 177 OT goals within 20 s; 5 against a defence 60 s+ in; York’s 96 s is the oldest defence on any of 148 5v5 OT goals (next 85.2 s); McCann’s 176 s the oldest on any regular-season 5v5 goal (next 173.4 s) | Scenes, not tested |
| Scorers specialise in tired defences | Split-half r of the stale-goal share -0.08 and of the post-icing share 0.15 (65 scorers); Auston Matthews 92 of 241 (38.2% vs league 33.6%, z 1.5); range among 109 100-goal scorers 23.2% (z -2.3) to 42.9% (z 2.0); Jake Guentzel 15 post-icing goals vs 5.6 at the league rate (11 of 95, then 4 of 75) | No (noise) |
Continues below
| Test | Result | Verdict |
|---|---|---|
| Centres specialise in stuck draws | Split-half r: stuck win % 0.56, other DZ win % 0.64, stuck minus other 0.12 (65 centres); best of 69 with 400+ stuck draws: Claude Giroux 63.3% (n 542) vs 62.6% on other DZ draws | No (faceoff skill, not stuck-draw skill) |
| Tired defences let in more goals than xG predicts | goals ÷ xG 1.00 at 40-50 s and 1.00 at 80 s+ pooled; above 1 only from 2023-24 | Not shown (unresolved split) |
| Coaches have cut how much play happens on tired legs | Stale share of attempts 29.0-31.8%, of time 18.1-19.3%, of xG 31.6-34.6%; mean defending age at a shot 32.4-33.8 s; no trend | No |
| Attempt rate against rises with shift age | 20.6 (under 10 s), 60.5 (40-50 s), 106.2 (80 s+) attempts per 60 | Descriptive only (zone confound) |
Continues below
| Test | Result | Verdict |
|---|---|---|
| Team timeouts at icings fell after the 2017 rule | 0.24 a game in 2009-17 vs 0.03 in 2017-26 (sampled games); since 2017-18, 1 of 1,864 icing stoppages with a TV timeout, 6 with a team timeout | Holds (sample of 24 games a season) |
Frequently asked questions
Do long shifts lead to goals against in the NHL?
Partly. 5-on-5 shots against a defence 80 seconds or more into its shift carry about 68% more expected goals (the chance a shot goes in, judged from where and how it was taken) than shots against a fresh defence. Matched to similar situations the gap shrinks to about 31% but stays, and part of it is teams being pinned in their zone, not only tired.
Why can’t a team change after icing the puck?
Since 2005-06, NHL rules (Rule 81.4) bar the team that ices the puck from changing before the next faceoff. Since 2017-18, 5-on-5 opponents create about 15% more expected goals in the 20 seconds after that draw than after other defensive-zone draws. The stuck team wins about 47% of those faceoffs.
Can a team call a timeout after icing the puck?
Not since 2017-18. Rule 87.1 says no timeout shall be granted to the team that ices the puck. The general managers recommended the change in March 2017 because coaches had been using their timeout to rest the stuck players. The other team can still call one.
Do tired defenders give up more goals than expected goals predict?
Not over the full nine seasons: goals ÷ expected goals is about 1.00 for defences 40-50 seconds in. From 2023-24 on, late-shift shots have beaten the model, but that break coincides with a change in how the NHL records shots, so we treat it as unresolved.
How long is a tired shift in hockey?
There is no hard line. In our data the danger rises steadily with time on the ice. The average 5-on-5 shot comes with the defenders about 33 seconds into their shift, and about 30% of shots come against defences 40 seconds or more in.
More in Minutes
- Verdict: Data story
The NHL Shift Got Shorter, Then Grew Back
- Verdict: Data story
Only 14 NHL Defencemen Still Score Like Ken Daneyko
- Verdict: Data story
The NHL Bench Grew From Eight Skaters a Night to 18 in Jumps
- Verdict: Data story
NHL Forwards Playing Under 10 Minutes a Night Fell From 43 to 16 Since 1998-99
Sources and method
- NHL API: game play-by-play and shift charts (api-web.nhle.com), 2017-18 to 2025-26
- NHL stats (NHL.com), every season since 1917-18
- Scouting the Refs: NHL historical rule changes
- NHL Official Rules 2024-25 (Rule 81.2, 81.4 and 87.1)
- NHL: 2017-18 rule changes (Rule 87.1, no time-out for the icing team)
- ESPN, Pierre LeBrun: GMs don’t want timeout after icing (March 8, 2017; Bettman, Cheveldayoff)
- Associated Press via Fox Sports: Cam York scores in OT as Flyers beat Penguins 1-0 in Game 6 (April 29, 2026)
- NHL play-by-play report: Pittsburgh at Philadelphia, Game 6, April 29, 2026
- NHL time-on-ice report (visitor): Pittsburgh at Philadelphia, Game 6, April 29, 2026
- ESPN, Emily Kaplan: Point gives Lightning win over Blue Jackets after 5 overtimes (Aug. 11, 2020)
- NHL play-by-play report: Columbus at Tampa Bay, Game 1, Aug. 11, 2020
- NHL time-on-ice report (visitor): Columbus at Tampa Bay, Game 1, Aug. 11, 2020
- Field Level Media via Sportsnaut: Monahan’s two goals lift Habs over Kraken (Dec. 4, 2023)
- NHL play-by-play report: Seattle at Montreal, Dec. 4, 2023
- NHL time-on-ice report (home): Seattle at Montreal, Dec. 4, 2023
- Hockey Analysis: are changes in NHL shot data tracking hiding better defending? (2025)
Research and analysis, not betting advice. Past results do not guarantee future returns; bet only what you can afford to lose.