Is the Rush Dying in the NHL?

By the scoresheet, the rush was dying

Through 2022-23, rush goals were a steady slice of the game: about one in every nine 5-on-5 goals, season after season. Then the slice shrank. By 2024-25 it was about one in thirteen, the lowest of the nine seasons, and it stayed low in 2025-26.

That count is ours. The NHL’s play-by-play logs every faceoff, hit, giveaway, takeaway and shot in every game, and from it we rebuild who had the puck and when it got it. If a team got the puck in its own half of the ice and scored within 10 seconds, we count the goal as a rush. Over nine regular seasons, 11,052 games and 44,542 5-on-5 goals, that rule tags 4,476 goals as rushes (10.0%), about one in ten.

Thompson’s second goal against Boston shows how strict that is. On the scoresheet, Buffalo got the puck 13 seconds before the goal, when a Sabre blocked a Boston shot in Buffalo’s end. The puck crossed the blue line only 5.8 seconds before it went in. But 13 is more than 10, so our count files the goal as something other than a rush.

The scoresheet’s story even has a champion. In June 2025 the Florida Panthers won a second straight Stanley Cup under Paul Maurice; the AP’s Stephen Whyno wrote that “the Panthers defeated Edmonton in six games in the final.” Sam Bennett credited the coach and a team that bought in to “play that hard style that he keeps preaching to us night in and night out.” Florida does score less off the rush, even on the puck-tracking replays. From 2023-24 to 2025-26, about 30% of its 5-on-5 goals were rushes on the replays, the third-lowest share of 31 teams. The league’s was about 35%. No team scored more goals by shooting the puck in and chasing it (22).

One champion can’t show that the whole league stopped rushing. And across the league, teams’ rush shares on the replays differ only a little more than luck alone would make them differ, so part of Florida’s low number may be luck too. To see the whole league, you need the league totals, and the chart above carries a warning.

Look at its biggest slice. For about a third of all goals, the play-by-play doesn’t record how the possession began at all. Our count can only see what the scorers write down. So before believing the rush is dying, we checked whether the game changed or the writing did.

The chances never left

Start with the shots. If the rush were really dying, teams would get fewer rush chances, and they didn’t. Using the same 10-second rule on shots instead of goals (shots on goal and misses, leaving out blocked shots), about one 5-on-5 shot in eleven came off the rush in every one of the nine seasons. That was 6.46 rush shots a game in 2017-18 and 5.87 in 2025-26, a small dip in a season when teams took fewer 5-on-5 shots of every kind.

Those rush shots didn’t get any easier to stop, either. Our shot model gives every shot a chance of going in from where and how it was taken, so a shot from the slot rates higher than one from the point. Add those chances up and you get expected goals. Rush shots carried about 7% of the league’s expected goals in 2017-18 and about 9% in 2025-26, while they stayed about one shot in eleven. A slightly smaller share of the shots carried a bigger share of the danger, so the model rated the average rush shot more dangerous by the end.

So, by our count, rush shots simply stopped going in. In 2017-18 about one rush shot in 15 became a goal (6.5%), and in 2021-22, the best of the nine seasons, about one in 12 (8.1%). By 2024-25 it was about one in 20 (5.0%), and in 2025-26 about one in 17 (5.9%). All the while, other 5-on-5 shots went in about once every 16 to 18 tries (5.5% to 6.2%). Teams were still getting rush shots just as often, and players don’t usually forget how to finish one kind of chance over three seasons. Scorers, though, can change how they write down the plays around it.

The worst season for rush shots wasn’t a recent one. It was 2020-21, the 56-game season, when about eight shots in 100 came off the rush. That season and the cut-short 2019-20 are counted as shares of the shots played, so the missing games don’t drag them down, and neither changes the picture.

Every way of drawing the line breaks in the same season

Maybe our line is in the wrong place. Thompson’s second goal missed the 10-second cut-off by three seconds. So we moved it. We tried 6 seconds and 15, and a stricter rule that needs a logged play in the shooter’s own half in the 4 seconds before the shot. Every version gives a lower rush share of goals in each season from 2023-24 on than in any season before it. The two strictest, 6 seconds and the 4-second rule, start slipping a season early, in 2022-23.

Every line steps down after 2022-23. The strictest one falls furthest: by the 4-second rule, about 3.5 goals in 100 were rushes in 2017-18, and fewer than one in 100 were by 2024-25. Our own 10-second line shows the shape most plainly. Through 2022-23 it drifted down by less than a tenth of a point a season (0.08). Measured across all nine seasons, the step-down after 2022-23 makes that 0.39 points a season. The fall sits in the break, not in a slow slide.

Only the goals broke. Run the same rules over every shot, most of which don’t go in, and the rush share barely moves. On the 10-second rule it was about nine shots in 100 in 2017-18 (9.4%) and about nine in 2025-26 (8.8%).

The scoresheet’s clock moved, too. Take the logged play that gave the scoring team the puck. In 2017-18, about 43 goals in 100 went in within 6 seconds of it (42.5%). By 2024-25 it was about 36 (36.0%). Saved and missed shots stayed between about 41 and 45 in 100 the whole time (41.4% to 45.3%). A shooter doesn’t know which shot will go in, so if teams had changed how they attack, you would expect the saved shots to move with the goals. They didn’t. Since 2023-24, goals sit further from the logged play that started their possession than they used to, and other shots don’t. That sends us to the people who log those plays.

In one summer, giveaways doubled and every rink fell in line

The writing did change. Our count builds every possession out of logged giveaways, takeaways, hits and shots, so how the scorers write those down decides what it sees. Two changes in that record-keeping arrived in the same seasons the rush share broke. In 2023-24 recorded misses jumped, which we covered in Shots on Goal Fell. Shooting Didn’t.. Then in 2024-25 recorded giveaways nearly doubled, from about 15 a game to about 30, while takeaways fell by about a third.

For seven seasons the giveaway line wanders between about 15 and 21 a game. Then it jumps, and it stayed up in 2025-26 at about 31 a game. Takeaways stayed down. Hits, there for scale, slipped only a little.

Players are very unlikely to have started losing the puck twice as often in one summer. The likelier reading is that scorers began logging some plays as giveaways that they used to log as takeaways, or didn’t log at all. We found no NHL announcement of a change. These counts have always been soft. In 2002, when the league dropped hits, giveaways, takeaways and blocked shots from players’ official summary lines, ESPN’s Terry Frei summed up its reasoning: the categories “involved subjective judgments that could vary from building to building.”

Each arena has its own stats crew, and it logs the giveaways and takeaways in that building’s games. In Columbus home games, the giveaways logged for both teams went from about 7 a game in 2023-24 to about 32 in 2024-25, nearly five times as many. Seattle went from about 7 to about 28, and Buffalo from about 9 to about 29. In St. Louis, takeaways fell from about 23 a game to about 9.

The buildings had disagreed wildly. A year earlier the range ran from about 7 giveaways a game (Columbus) to about 25 (Montreal). The Hockey Writers’ Daniel Amoia noticed in September 2024: “The Calgary Flames were dead-last in the NHL last season with 532 giveaways at home.” And: “The Columbus Blue Jackets, on the other hand, were the best in the NHL with 151 giveaways in their barn.” His complaint was that the stat “leaves it up for the home scorer’s interpretation,” meaning the arena’s own stats crew.

Amoia’s 532 and 151 are totals for a whole season, and he counted a different way from us. We add both teams’ giveaways and divide by the number of games. His top building is Calgary, and ours is Montreal. Both counts have Columbus logging the fewest.

In 2024-25 the disagreement all but vanished. Every rink logged between about 28 giveaways a game (Seattle) and about 33 (Carolina). Crews in 31 buildings don’t drift to the same number in one summer by chance. It looks like a league-wide change in how turnovers are logged, and it arrived in the season our rush count sank.

The replays saw what the scoresheet missed

If the scoresheet changed, the way to check is a record nobody types in. Since 2023-24 the NHL has tracked the puck and every player on the ice, and we went through those replays for nearly every 5-on-5 goal. They mark each zone entry, the moment the puck crosses the other team’s blue line, so they don’t depend on anything a scorer chose to write.

On the replays, we call a goal a rush when it went in within six seconds of the zone entry. Deflections, rebounds, faceoff plays and dump-ins (pucks shot in and chased) don’t count. In each of the three replay seasons, 2023-24 to 2025-26, about 35% of 5-on-5 goals were rushes. Those are the same three seasons in which our scoresheet count sat at its lowest.

The two records were never going to agree on how common the rush is. Most real rushes start with a play the scorer never logs, a breakout pass or a loose puck picked up in stride. When our scoresheet count does call a goal a rush, it is usually right: about four times in five, the replays agree (80.5% in 2024-25). But it misses most rushes, and in 2024-25 it missed more. It caught about one replay rush in five in 2023-24 and about one in six a year later.

Those misses are most of the scoresheet’s drop. Take every 1,000 goals. In 2023-24 our count called about 95 of them rushes, and about 76 of those were rushes on the replays too. In 2024-25 it called about 77, and about 62 were replay rushes. The replays had about as many rushes both years. So our count made about 18 fewer rush calls per 1,000 goals, and about 14 of those were real rushes it stopped catching.

Takeaways show where some of those misses came from. Our count leans on a logged takeaway: if the scoresheet says a team stole the puck in its own half and shot within 10 seconds, the count calls it a rush. In 2023-24 the replays showed about 80 rushes that, on the scoresheet, began with a takeaway in the team’s own half. Our count caught nearly all of them. In 2024-25, the season the rinks logged about a third fewer takeaways, the scoresheet had only about 40 replay rushes starting that way, and our count caught about 30. Some of those starts were logged as giveaways instead.

Add up every kind of start and our count caught 73 fewer replay rushes in 2024-25 than a year earlier. Turnovers explain only part of that. It caught 49 fewer rushes that began with a takeaway and 22 more that began with a giveaway. That leaves turnovers 27 of the 73, a little over a third.

Plays that began some other way lost more: about two thirds of the loss on their own. Those are mostly a team picking up the puck after the other team’s shot, like the Edmonton pickup before Leon Draisaitl’s 2-on-1 below, or a hit. Faceoff starts won a few back. The replays had about as many rushes that began with a recovery or a hit as the year before. But in 2024-25 the scoresheet more often put their start more than 10 seconds before the goal, too long ago for our count. The giveaway jump is still the clearest sign the scorers changed their habits. What explains the rest of the missed rushes, the data can’t say.

The rush is still there, though the evidence has gaps

Put it together and the rush is still there, as far as the shot counts and the replays show. Teams got rush chances as often as ever. The replays show rush goals holding steady from 2023-24 on. Only one count collapsed, and it is the one built from the scoresheet. It stopped catching rushes the replays still see. That happened in the season the scorers changed how they log turnovers, but the turnover change explains less than half of the rushes it missed.

We are surest about that break in the scoresheet. We are less sure about the years before the replays began, because the one outside count we found points the other way. In January 2025, Michael Clifford of Dobber Hockey reported numbers from AllThreeZones, an independent project whose trackers watch games and log plays by hand: “In 2022-23, the average forward managed 9.23 shots or assists on teammate shots off the rush per 60 minutes.” Give a typical forward 60 minutes of 5-on-5 ice time, a full game’s worth, and in 2022-23 he took or set up about nine rush shots. A season later it was about eight (8.01), and in 2024-25 the figure was “landing at 7.94 as of Wednesday afternoon.” Those counts don’t come from the arena stats crews, and they fell.

Their count and ours measure different things. AllThreeZones counts each forward’s rush shots and setups, in the games its trackers watched. We count the share of all shots that our scoresheet count calls rushes: about 9% in 2022-23, 10% in 2023-24 and 9% in 2024-25. We can’t see their method or which games they tracked, so we can’t line their drop up against our numbers.

What we can say is when it happened. Most of their drop came between 2022-23 and 2023-24, and the replays begin in 2023-24, so they can’t check it.

Our own evidence has gaps too. Because the replays only cover 2023-24 on, we can’t check the scoresheet against them in earlier seasons. Our count was flat from 2017-18 to 2022-23, but that is the scoresheet’s view, not the rink’s.

The replays may be drifting a little. The quickest goals, within 6 seconds of the puck crossing the blue line, held steady from 2023-24 to 2025-26. Goals a few seconds slower, 7 to 10 seconds after the entry, became more common: all goals within 10 seconds went from about 51% to about 58%. We can’t tell whether that is the game or the tracking.

Turnovers aren’t the whole story either. Leave out every possession that began with a takeaway or a giveaway, and our count’s rush share of goals still falls, from 7.0% in 2017-18 to 4.7% in 2025-26. So something besides turnovers changed too. Since 2023-24 the replays say it wasn’t the rush itself. Before that, we can’t tell whether it was the logging or the game.

We found no NHL explanation for the 2024-25 jump in giveaways. Every rink landing on about the same count points hard at a new way of logging giveaways, but nobody has confirmed one. And the players named here show what the scoresheet misses. One player’s season is a few dozen 5-on-5 goals at most, few enough that luck can move the count, so treat them as examples of what the scoresheet misses in a single season.

A falling rush stat may just be the scoresheet

That matters the next time a stat site or a broadcast graphic says the rush is fading. If the number is built from the play-by-play, be wary of any trend that runs across 2023-24 and 2024-25. Any count that leans on giveaways, takeaways or the timing of logged plays breaks between those two seasons, and a drop there may say more about the scorers than about the players.

Rush numbers from puck tracking, like the replays, are the better guide from here, because the moment the puck crosses the blue line doesn’t depend on a scorer’s judgement. Three seasons of the replays show no decline.

Thompson was moving his feet

Back in Buffalo that January night, the Sabres won 7-2. Lindy Ruff said: “There’s a lot of different things we’ve tried that I think we’re finally getting a benefit from.” Thompson put it in fewer words. “We’re at our best when we’re moving our feet,” he said.

The scoresheet from that night lists three goals by Thompson and no rushes. The replay has the puck carried over the blue line and into the net inside six seconds, three times.

Show the work

Tests written down

Not in advance (exploratory)

Results computed

October 27, 2026

Every 5-on-5 goal and unblocked attempt (both goalies in net, shootouts excluded) in 11,052 regular-season games from 2017-18 to 2025-26, from our play-by-play database. Possession chains are rebuilt from recorded events: a chain starts with a faceoff win, takeaway, opponent giveaway, hit taken or recovered opposing shot, and ends on a change of team or 20 seconds with nothing logged. Play-by-play rush (the article’s “our count”) = possession began in the neutral or defensive zone, shot within 10 s; across the nine seasons that rule tags 4,476 of 44,542 5-on-5 goals. Rush share of 5-on-5 goals: 10.0% to 11.7% each season through 2022-23, 9.4% in 2023-24, 7.7% in 2024-25 (lowest of nine), 8.5% in 2025-26. Rush shots were 8.2% to 9.8% of unblocked attempts every season (6.46 a game in 2017-18, 5.87 in 2025-26); their expected-goal share rose from 7.4% to 8.7% while their share of unblocked attempts went from 9.4% to 8.8%, so expected goals per rush shot rose from 0.0447 to 0.0606 (other shots 0.0577 to 0.0617). Our expected-goals model takes the play-by-play rush flag, the 4-second rule and the seconds since the possession began as inputs, so the rush expected-goal share is not independent of the rush count; read it as “rush shots were not rated less dangerous,” not as an outside check.

Continues below

Rush shooting percentage was 6.5% in 2017-18, peaked at 8.1% in 2021-22, slid to 5.0% in 2024-25 and was 5.9% in 2025-26; every other 5-on-5 shot went in at 5.5% to 6.2%. Replay rush (2023-24 on; the article’s “the replays”) = our analysis of NHL puck and player tracking on goal replays: zone entry to release within 6 s, after deflection, rebound, faceoff and dump-in checks. Replay rush share 34.5%, 35.2%, 35.4%; goals within 6 s of any entry 44.1%, 44.7%, 45.0%; within 10 s 50.8% to 57.6%. When the play-by-play count calls a goal a rush, the replay agrees 80.5% of the time (2024-25); it caught 21.9% of replay rushes in 2023-24 and 17.6% in 2024-25. Checks: 6 s and 15 s windows, a 4-second rule, the same rules on attempts, a version without turnover-started possessions, event recording counts league-wide and by home rink, the replay label for 2023-24 to 2025-26 (and which replay rushes the play-by-play count caught, by how the logged possession began), team splits by season pair, and team speed from NHL EDGE. Player lines and named goals come from the same goal rows; season goal totals from the NHL stats API. Arizona and Utah are left out of the rink and replay team comparisons. Definition checks.

Continues below

Windows of 6, 10 and 15 s from a possession start in the scoring team’s neutral or defensive zone, and the 4-second rule: any recorded event in the shooter’s neutral or defensive zone in the 4 s before the shot, with no stoppage between. Goal share 2017-18 vs 2024-25: within 15 s 18.1% vs 15.3%; 10 s 10.9% vs 7.7%; 6 s 4.6% vs 1.4%; 4-second rule 3.5% vs 0.7%. The same rules on unblocked attempts: the 10-second share was 9.4% in 2017-18 and 8.8% in 2025-26. Goals within 6 s of the logged start of their possession: 42.5% in 2017-18, 36.0% in 2024-25; saved and missed shots 41.4% to 45.3% in every season. Without turnover-started possessions the 10-second goal share is 7.0% (2017-18) and 4.7% (2025-26). Event recording, exact: recorded giveaways, both teams, all strengths, 15.4 a game in 2023-24 and 30.0 in 2024-25 (30.8 in 2025-26); takeaways 14.0 to 9.6 (9.2); hits, for scale, 45.5 to 42.9. By home rink: Columbus 6.6 to 32.4 giveaways a game, Seattle 6.8 to 27.5, Buffalo 8.6 to 29.4; St. Louis takeaways 22.9 to 9.0. Rink range 6.6 (Columbus) to 25.4 in 2023-24; 27.5 (Seattle) to 32.7 (Carolina) in 2024-25. Our attempt-share measure, for comparison with AllThreeZones: 9.1% in 2022-23, 9.8% in 2023-24, 9.0% in 2024-25.

Continues below

AllThreeZones counts each forward’s rush shots and shot assists (passes to a teammate’s shot) per 60 minutes of 5-on-5 play. Takeaway-started rushes, exact counts (own half = the scoring team’s neutral or defensive zone). The play-by-play count caught 76 of the 81 replay rushes whose logged possession began with an own-half takeaway in 2023-24, and 27 of 36 in 2024-25. It caught 22 more rushes that began with an own-half giveaway, so turnover starts account for a net 27 of the 73 fewer replay rushes it caught (396 to 323). Own-half recovered-shot and hit starts lost 50 (their replay rushes 626 to 652); faceoff starts gained 4. Florida, replay data 2023-24 to 2025-26: 30.0% of 5-on-5 goals were replay rushes, third-lowest of 31 teams (league 35.0%); 22 dump-and-chase goals, the most. Team spread of replay rush shares: standard deviation 2.89 points observed against 2.16 expected from chance alone. Team habit and speed, method. Team spread is the standard deviation of team rush shares across the 31 teams with seven or more seasons, against the spread chance alone would give: 0.52 points against 0.19 for unblocked attempts, 0.95 against 0.81 for goals. Season-to-season repeat: r = 0.33 for attempts (251 season pairs, 95% range 0.21 to 0.43) and r = 0.16 for goals, about 0.05 once each season’s league-wide average is subtracted from every team.

Continues below

Frequently asked questions

More in The Vanishing Game

• Verdict: Data story

  The Last Goalie Who Played Every Night

• Verdict: Data story

  Did Garbage Goals Take Over the NHL?

• Verdict: Data story

  For 19 Seasons, an NHL Hit Depended on the Rink

• Verdict: Data story

  Hainsworth’s 22 Was Mostly Arithmetic

• Verdict: Data story

  The Slapshot Went From One NHL Goal in Four to One in Twelve

• Verdict: Data story

  NHL Fights Fell to 0.18 a Game in 2018-19, Then Climbed Past the 0.20 of Gordie Howe’s Era

All How Hockey Changed research →

Keep reading

• How Hockey Changed: NHL data stories →One chart-led story a day about how the NHL game, its players and its rules have changed.
• Shots on Goal Fell. Shooting Didn’t. →The 2023-24 break in how NHL misses and shots are recorded.
• Does skating speed win games? →A lineup speed rating tested against expected goals, wins and closing odds.
• Arena scorer bias →How the home rink’s scorer changes counted stats like hits.

Sources and method

1. NHL stats (NHL.com), every season since 1917-18
2. NHL API reference (community documentation of the stats endpoints)
3. The Hockey Writers, Daniel Amoia: why NHL giveaway and takeaway stats shouldn’t be trusted (Sept. 8, 2024)
4. NHL.com, Heather Engel: Bruins at Sabres recap (Thompson hat trick, Jan. 28, 2025)
5. AP via FOX 32 Chicago: Avalanche 5, Blackhawks 0 (MacKinnon’s 100th point, Feb. 29, 2024)
6. ESPN, Terry Frei: NHL taking the crunch out of numbers (2002)
7. Dobber Hockey, Michael Clifford: Ramblings, Jan. 16, 2025 (AllThreeZones rush data)
8. AP via NBC Chicago, Stephen Whyno: Paul Maurice and the Panthers repeat (June 17, 2025)

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