How Much Do NHL Referees Change the Power-Play Count? The Busiest Crews Call About 14% More Than the Quietest

Tim Peel’s games had an ordinary number of power plays

The league’s statement came from Colin Campbell, its senior executive vice president of hockey operations. “There is no justification for his comments, no matter the context or his intention,” Campbell said. The league punished what he said. What his whistle did over four seasons is something we can count.

In the 11,052 regular-season games from 2017-18 to 2025-26, the two teams combined for about 5.9 power plays a game. In all, 60 referees worked at least one of them. For every game we worked out how many power plays the two teams, the season and the month should produce. Then we set each referee’s games against those predictions. Peel’s 207 games ran about 0.1 of a power play a game above (+0.12), well inside what luck alone produces for a referee with that many games. Two referees work every game, so that number mixes Peel with his partners. Split the credit for each game between the pair, and Peel’s own number is +0.05 a game. It could plausibly sit anywhere from −0.17 to +0.28, a range that includes zero. None of this tests the Arvidsson call itself. A single call in a single game can’t be tested this way.

Luck alone moves these counts around. A referee who works 50 games can have a run of chippy nights and look busy. Over 500 games those nights wash out. How far a referee’s number usually drifts by luck alone is his wobble. For a referee with 200 games, it is about 0.15 of a power play a game. The grey lines in the chart below sit about two wobbles from the prediction (the inner pair) and about three (the outer pair), and they narrow as the games pile up. If referees did not matter at all, 95 in 100 would land inside the inner pair and almost every dot would sit inside the outer pair. The chart shows the 54 referees with 50 or more games.

Peel sits in the crowd. Plenty of other referees don’t. Of the 47 referees with at least 100 games, 12 are further from average than luck alone would almost ever put them. If referees did not differ, the odds would be better than nine in ten that none of the 47 landed out there (the expected count is about 0.1). Of the same 47, 19 sit beyond the inner grey line, where chance would put two or three (about 2.4). So referees really do differ. Over these nine seasons, a typical referee’s games have about a third of a power play a game more or fewer than average (0.33). That is about 6% of the league’s 5.9 power plays a game.

The players’ favourite referee had one of the quietest whistles

So who sits out at the edges? One of the quietest is the referee players like best. In the spring of 2018 the NHL Players’ Association asked players to name the league’s best referee. “McCauley earned the votes of 47.8 percent of players as the NHL’s best referee, well ahead of Kelly Sutherland (17.7 percent),” the Associated Press reported. Peel got 4.4 percent. Wes McCauley is the showman of the job. “When he reaches to his right hip to turn on the microphone, it’s must-see entertainment,” the AP wrote. His own goal was plainer: “to do the best I can and to move on to the next game and really to stay out of the highlights.”

McCauley’s games had fewer power plays than almost anyone’s. Over 632 games, the teams got about 0.6 fewer a game than their matchups predicted (−0.59). Luck alone keeps a referee with that many games within about 0.26 of the prediction 998 times in 1,000. McCauley is more than twice that far out, about seven wobbles. With the credit split between partners he comes out at −0.57 a game, probably between −0.70 and −0.43.

A harder test is to rate each referee on 2017-18 to 2022-23 only, then watch the games that came after. On those earlier seasons McCauley’s games ran about 0.7 a game below the prediction, over 406 games. In the 226 games since, they ran about 0.3 below. Luck’s wobble over 226 games is about 0.14, so he sat about two wobbles under. That is still bigger than luck would usually produce. His habit probably carried into the new games, at less than half its earlier size.

Players and coaches said much the same. “Consistency. You kind of know what you’re getting with him,” said Chris Kreider, then with the Rangers. “He stands his line and he lets the players play,” said Todd Reirden, then Washington’s coach. The power-play count says the same thing.

Michael Markovic is the other end. His 393 games ran about 0.8 power plays a game above the prediction (+0.78). The outer luck line sits about three wobbles out. Markovic sits about seven (7.16), further from the prediction, measured in wobbles, than any other referee. With the credit split between partners he is at +0.78, probably between +0.61 and +0.94. He was above in all five of his seasons with 40 or more games. Rated on 2017-23, his games ran about 0.9 a game above, over 171 games. In the 222 games since, they ran about 0.6 above, about four wobbles. That is far too big to be luck, and the highest of the 31 referees in that test.

Split the credit between partners, though, and Cody Beach has the top number of all, about 0.8 a game over 212 games, probably between +0.58 and +1.02. That is too big to be luck. But his first game in our data was April 12, 2022, and he worked too few games before 2023-24 for the test on later games.

The chart below shows the regulars: the 43 referees who worked 200 or more games (the 47 above counted everyone with 100 or more). Two referees work every game, so the chart splits the credit for each game between the pair. A referee with only a few games gets a cautious number, nudged toward average, because a short run of games can mislead. Cody Beach (212 games) and Michael Markovic (393) lead. Dean Morton (335) and Wes McCauley (632) are at the other end.

A Cody Beach game would have about 6.7 power plays where a Dean Morton game would have 5.2. So the Beach game has about 28% more power plays than the Morton game, well short of double. On new games, only about half of a gap like that has carried over. Crews matter a little more than one referee, because the two referees’ habits add up. Even so, in nine games out of ten the pair moved the count by less than one power play either way, somewhere between −0.62 and +0.73.

Most referees are close to average: 28 of the 43 regulars are within a quarter of a power play either way. Only 19 of the 43, with the credit split between partners, are far enough out that the gap is probably real. Most of them are the same men as the 19 beyond the inner grey line in the first chart: 18 of the 19 are on both lists. None of this grades a call. The rulebook leaves room for judgement, and both ends of the chart work inside it.

A referee’s habit lasts, then drifts

If those numbers were only luck, they would not repeat. Mostly they do. Split each referee’s games into two piles, odd-numbered and even-numbered, and the two piles look much alike. Score the match from 0 (no link at all) to 1 (the piles line up perfectly), and the 43 referees with 200 or more games come out at about 0.8 (0.81). One season against the next is a harder test, because a season is only about 67 games. On the same scale, 248 pairs of back-to-back seasons match at 0.54. A referee whose games ran one power play above average in one season ran about 0.5 above the next. Chris Schlenker’s games were above the prediction in all eight of his seasons with 40 or more games, Chris Lee’s below in all eight, and McCauley’s below in eight of nine.

That 0.54 is a little low. Even if every referee kept exactly the same habit, luck in a 67-game season would blur the match. Our numbers say back-to-back seasons would then score about 0.61. On its own, the shortfall could be luck, because the same referees turn up in many of the pairs. But the test on later games, further down, points the same way. Habits drift.

Dean Morton’s drifted the most. In 2017-18 his games ran 0.20 a game above the prediction. In 2020-21 they averaged 4.04 power plays, when the league averaged 5.78. In 2021-22 they averaged 4.21. Count every season in which a referee worked 40 or more games; there are 307 of these referee-seasons. Morton’s last two are the quietest of them all, about 1.5 and 1.6 power plays a game below the prediction. The next quietest, Marc Joannette’s 2020-21, was 0.9 below. Across all 335 of his games, Morton has the lowest number of the 43 regulars, about 0.7 a game below average, probably between −0.84 and −0.48.

Could luck alone hand one referee the two quietest seasons? If referees did not differ at all, the two quietest of 307 seasons would belong to the same man about 2% of the time. Referees do differ, and a quiet referee is more likely to have two quiet years. So we had a computer replay the nine seasons 20,000 times, with referees as different from each other as they really are. In those runs the two quietest seasons landed on one referee about 27% of the time. That part is not rare. What the runs almost never produced was a second season as quiet as Morton’s. It turned up in only about one run in 200 (0.5%). One man with two seasons that quiet came up in 0.4%. His first seasons were near average, so this looks like a change late in his career. Neither our data nor any source we found says why. 2020-21 was a short season with few or no fans, but every referee worked it, and none came near his numbers.

Morton also turns up in the strangest games of these nine seasons. Four of the 18 games with no power plays at all had him on the ice: March 13, 2021, and three in his last season, on Oct. 19, 2021, Jan. 24 and Feb. 27, 2022. His share of all games would give him about half of one (0.55). For him alone, four or more would come up about 0.24% of the time, roughly one chance in 400. It looks damning, but it is the kind of thing luck does. With 60 referees, the chance that at least one of them reaches four is about 39%, and Garrett Rank also had four. The two quiet seasons are the evidence; the empty games are just a memorable detail.

His last game came on March 12, 2022, Detroit at Calgary. FlamesNation called it fitting. Morton had played exactly one NHL game, for Detroit, at the Saddledome in 1989-90, and scored in a 10-7 loss. The goal was “assisted by Hall-of-Famer Steve Yzerman,” Scouting the Refs noted. The 2022 game had 5 power plays, 3 for Calgary and 2 for Detroit, and Calgary won 3-0. After the final buzzer, FlamesNation wrote, “the players from both teams lined up to shake his hand and wish him the best.” By its count it was his 1,005th game.

On new games, the habits kept a little under half their size

McCauley’s habit probably carried into new games and Markovic’s clearly did, both at a smaller size. To test every referee the same way, we rated each one on 2017-18 to 2022-23 only. Then we looked at the 3,936 games of 2023-24 to 2025-26, which the ratings had never seen. We kept the 2,733 of them where both referees had worked at least 100 games in the earlier seasons, about 69% of the total, so that each crew’s rating rested on a real record. We gave each of those games its crew’s earlier rating and split them into five equal piles, from the busiest-rated crews to the quietest. The busiest fifth had about 6.2 power plays a game (6.20) and the quietest fifth about 5.4 (5.44), while the teams involved predicted almost the same number for both. That is a gap of about three quarters of a power play a game (0.76). Luck alone could move each pile by about a tenth either way, so the gap is far bigger than chance. So games with the crews rated busiest had about 14% more power plays than games with the crews rated quietest. That is smaller than the 28% between a Beach game and a Morton game. The piles mix many crews, while Beach and Morton are the two extremes, and the piles use only games the ratings had never seen.

So the habits McCauley and Markovic showed were not flukes, and they did shrink. Take the 31 referees who worked 100 or more games in both periods. A referee’s earlier number tells you a good deal about his later one: on the same 0-to-1 scale as before, the two line up at about 0.6 (0.60). But on average only about 45% of his old habit showed up again. Try to guess each game’s power plays. A guess that knows the crew misses by almost exactly as much as one that doesn’t: it is 0.3% better. Most of the variation is the game itself. One night in Florida shows how much.

On Dec. 27, 2025, McCauley and Sutherland, the players’ top two picks in 2018, worked Tampa Bay at Florida. Both are quieter than average in our data: McCauley we have met, and Sutherland’s 601 games ran about a quarter of a power play a game below (−0.25, probably between −0.39 and −0.11), probably real but small. So the two referees’ records pointed to a slightly quieter night than usual. The game had 16 power plays, 10 for Florida and 6 for Tampa Bay, the second-most of the 11,052 games. Only Chicago against Edmonton on Oct. 27, 2022, with Chris Rooney and Jon McIsaac, had more (17). Late in the second period a cross-check by Anthony Cirelli on A.J. Greer “ushered in a multi-player brawl between both teams,” the Panthers’ website reported, and the night finished with 136 penalty minutes. Florida went 1-for-10 on the power play and lost 4-2. “We’re kind of used to that with these guys,” said Paul Maurice, the Florida coach. Nothing in the two referees’ records pointed to a night like that.

Late in a close game, calls fall by about half, whoever the referee

If the crew is worth less than one power play, the clock is worth more. In a tied or one-goal game, referees make about 6.7 penalty calls per 60 minutes from minute 25 to minute 40. At that pace, a whole game played like that stretch would have about seven calls. Calls dip at the start of the third period (5.0) and keep falling. In the last five minutes of a tied or one-goal game they are down to 3.5, about half. That last stretch covers 28,016 game-minutes (minutes of play added up across all those games) and 1,628 calls, so it is no small sample. With two or more goals between the teams, the third period is quieter too, but the last five minutes only fall to 4.7.

One check before going on, because the clock counts calls and this story is about power plays. The two are close, but not quite the same. Over all of regulation we count 62,245 calls in 663,120 game-minutes, about 5.6 calls a game (5.63), a little under the 5.9 power plays (5.86). Most of the gap comes from scrums. When players from both teams go to the box at the same second, our call count skips that second. But if one of them got four minutes and the other two, one team still gets a power play. The power-play count goes up and the call count doesn’t. Power plays in overtime are not in our calls either. When we checked 2017-18 and 2025-26 game by game, about a quarter of the gap was overtime and about two thirds came in games with one of those both-team moments. A few calls push the other way. A misconduct on its own sends a player off for 10 minutes but leaves the teams at full strength, so it counts as a call and gives no power play.

Peel said on the microphone that he wanted a penalty “early.” Early is already a quiet time. In a close game, calls run about 4.2 per 60 minutes in the first five minutes and 5.8 in the first five of the second period, when the Arvidsson call came. Only the last five minutes of a game are quieter than the opening five.

This is a habit of the whole league, measured across every crew at once. The count cannot say who is holding back. Referees may let more go late, players may play more carefully with a game on the line, or both. We found no referee or league official on the record explaining the drop.

The score steers the whistle too, and the next call tends to go to the other team

The clock is not the only part of the game that steers the whistle. The score does too. A team leading by one is called about 3.1 times per 60 minutes, against about 2.6 for the team trailing by one and 2.7 for each team when tied. These count each team separately, so they run about half the clock figures above, which counted both teams together. The quietest spot in the score count is a tie late on. In the last ten minutes of a tied game, each team is called only about 1.8 times per 60.

Referees may also shape a game’s power plays by spreading them out. The calls tend to take turns. Across 51,212 pairs of back-to-back calls, the next one went to the other team about 59% of the time (58.9%). To see what chance would do, we shuffled each game’s calls into a random order, keeping each team’s number of calls. The shuffled games hand the next call to the other team about 55% of the time (55.2%). That is a bit more than half because a game has only a handful of calls. Picture a game where each team takes three penalties. Once Detroit has taken its first, five calls are left: two for Detroit and three for Nashville. So a random order hands the next one to Nashville three times in five. The real games beat that chance figure in all nine seasons. That fits referees evening things up, calling a penalty on the team that has just had a power play. But players change their behaviour with the score and the clock too, so this cannot say who is causing it.

The night in Nashville shows what alternation looks like on a scoresheet. Peel worked it with Kelly Sutherland. Detroit’s Adam Erne was called at 13:41 of the first period, with Arvidsson listed as the player who drew it. Arvidsson’s tripping call, at 4:56 of the second, was the game’s second. From the first call on, they went Detroit, Nashville, Detroit, Nashville, then a pair of coincidental penalties (one to each team at once, so no power play), which our counts leave out, then Nashville. It shows the pattern on one night and explains none of the calls. Two referees were on the ice, and nothing we have says which one made each call or why, beyond Peel’s own words.

Tonight’s crew is worth less than one power play

So for anyone checking the referees before puck drop, the crew is a small thing next to the clock and the score.

• The crew’s range is narrow. In nine games out of ten the two referees together moved the count by between −0.6 and +0.7 power plays, on a base of about 5.9. Going forward, expect about half of a referee’s past number to show up.
• We found no referee who favours the home team in a way that repeats. The home team gets about 52% of power plays league-wide. Referees differ in that share by about what chance produces. A referee’s home lean in one half of his games says almost nothing about the other half: on the 0-to-1 scale, the two halves match at 0.08.
• Referees do not measurably move the score. Earlier research on this site found that over 2,618 games with closing lines, goals against the market in each referee’s games were no more spread out than chance.
• The league trend is a separate story. Power plays fell from 11.7 a game in 2005-06 to 5.41 in 2024-25 across every crew; the referee differences here are measured inside each season, so they do not explain that slide. See the power play is disappearing.
• Tonight’s crews and their power-play numbers are on NHL referees tonight.

The night Peel lost his job was an ordinary night

Peel never finished the sentence. The audio cut out after “early in the …”, and the NHL finished his career for him the next morning.

The game in Nashville had 5 power plays, 3 for Detroit and 2 for Nashville. That is about an average night. His 207 games sit near the middle of the first chart, in the crowd of referees whose games look like everyone else’s.

Show the work

Tests written down

Not in advance (exploratory)

Results computed

October 27, 2026

11,052 regular-season games, 2017-18 to 2025-26. Power plays per team-game are modelled from season-month, home ice, team-season drawing and giving terms, and both referees’ terms, all shrunk as random effects. Two referees work every game, so every game is credited to the pair and each referee’s share is estimated alongside the team and season terms. Each estimate is then shrunk toward average by about 41 games’ worth of “average referee”, so a referee with 200 games keeps most of the raw number and one with 50 keeps about half. The funnel uses the same model without referees; its limits are ±1.96 and ±3.09 standard errors (per-game SD 2.15 ÷ √games). Repeatability: odd/even games and consecutive seasons. Holdout: fit 2017-18 to 2022-23, test 2023-24 to 2025-26; a named referee’s holdout standard error is 2.15 ÷ √(test games). Referee-seasons: every referee with 40+ games in a season (307), raw power plays against the no-referee expectation. The Morton benchmark simulates each referee’s true effect from a normal spread of 0.33 plus season noise (2.15 ÷ √games), 20,000 runs. The empty-game benchmark gives each referee a Poisson count with mean (his games ÷ all games) × 18. Penalty calls for the clock and score figures: play-by-play penalty events in regulation, one per team per second, leaving out coincidental penalties.

Continues below

Luck lines (first chart): for a referee with n games, chance alone keeps his raw number within ±1.96 × 2.15 ÷ √n of the prediction 95% of the time and within ±3.09 × 2.15 ÷ √n 99.8% of the time; 2.15 is the per-game standard deviation of power plays around the prediction, and 2.15 ÷ √n is the “wobble” (one standard error) in the body. Chance counts outside the lines are the 47 referees with 100+ games × 0.05 and × 0.002. Labelled points are referees more than five standard errors from expected, plus Tim Peel. “Wobbles” for McCauley and Markovic are their funnel z-scores (−6.86 and +7.16). The typical-referee figure (0.33) is the standard deviation of true referee effects in the joint model. Partner-separated numbers (bar chart, “probably between”): each referee is pulled toward average by about 41 games’ worth of an average referee; “probably between” is the 90% interval, and “far enough out that the gap is probably real” means that interval excludes zero. The crew range is the 5th to 95th percentile of the two referees’ combined effect across games. The 0-to-1 scale in the body is the Pearson correlation: odd against even games, one season against the next (about 67 games a season), earlier against later rating in the holdout, and home share in odd against even games. The season-to-season trend line and the holdout carry-over (45%) are least-squares slopes.

Continues below

The model-implied season-to-season figure assumes fixed referee effects with the model’s spread and each season’s sampling noise; the same referees appear in many pairs, so the gap between it and the real figure is within the noise on its own. Holdout: a named referee’s wobble on later games is 2.15 ÷ √(test games). Each fifth of crews carries a chance error of about ±0.1 power plays (one standard error). “Barely improves a guess” is the cut in mean squared error of a game’s power-play count from adding the crew’s 2017-23 rating to the team-season-month prediction. Morton’s seasons: “if referees did not differ” gives every referee-season the same chance of being lowest; the 20,000-run replay draws each referee’s true effect from the model’s spread. Empty games: the chance for him alone is a Poisson tail at his expected count. Evening up: consecutive pairs of calls within a game; the shuffle permutes each game’s calls and keeps each team’s count, which is why chance sits above 50%. Calls against power plays: the calls leave out every second in which both teams were penalised, including the ones where unequal penalties still gave one team a power play, and they leave out overtime. The quarter-and-two-thirds split of the gap comes from an audit check of 2017-18 and 2025-26 against the play-by-play; lone misconducts push the other way.

Continues below

Frequently asked questions

More in Rules

• Verdict: Data story

  A Shootout Shooter Needs More Than 200 Tries Before Skill Outweighs Luck. The Most Anyone Has Taken Is 135.

• Verdict: Data story

  Does an NHL Game Really Stop Less Than It Used To?

• Verdict: Data story

  The Overtime Faceoff’s 30-Second Head Start

• Verdict: Data story

  Spearing Hit Zero. Hooking Never Went Away.

• Verdict: Data story

  NHL Scoring Doubled in 1929-30 Under the New Forward-Pass Rules

• Verdict: Data story

  NHL Overtime Has Risen From 18% of Games to 23% Since 1999

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.
• The power play is disappearing →NHL power plays fell by half since 2005-06; scoring came back at even strength.
• NHL referees →Tonight’s officials and how many penalties each crew calls.
• Power plays tonight →Which teams are most likely to get power-play chances tonight.

Sources and method

1. NHL stats (NHL.com), every season since 1917-18
2. NHL stats API: power-play opportunities per team per game (team/powerplay game reports)
3. NHL game centre: the referees for every game (api-web.nhle.com right rail)
4. NHL play-by-play feed: penalty and goal events (api-web.nhle.com)
5. Scouting the Refs: NHL historical rule changes
6. Detroit Free Press, Helene St. James, via Yahoo Sports: NHL referee’s hot-mic call during Detroit Red Wings game ends his career (March 24, 2021)
7. Associated Press, Stephen Whyno, via ESPN: NHL’s best referee, Wes McCauley goes by his own book (Sept. 19, 2018)
8. Scouting the Refs, Josh Smith: NHL referee Dean Morton calls it a career (March 12, 2022)
9. FlamesNation, Mike Gould: Flames, Red Wings players shake hands with referee Dean Morton after his final NHL game (March 13, 2022)
10. FloridaPanthers.com, Jameson Olive: Panthers return from holiday break with loss to rival Lightning (Dec. 28, 2025)

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