HGB Research
The price of a penalty
The average two-minute penalty costs 3.3% of a win (2.2% in the last 2:20 of a period). Tied late in the third period, it costs 8–10%. We priced 47,909 regulation minor penalties that created a power play, 2018-19 to 2025-26, by the score, period and time they were taken.
The average two-minute penalty costs the penalized team 3.3% of a win. Penalties in the last 2:20 of a period are priced separately and average 2.2%. On average, a team at 50% to win drops to about 46.7%.
A hook in the first period of a 4-1 game and a hook with five minutes left in a tie put the same power play on the ice. The second one costs more, because a single goal decides more of the result when the game is close and the clock is short. So we measured penalties by their effect on win probability, not just by goals. ”% of a win” below means percentage points of win probability.
The average penalty: 3.3% of a win
We looked at 42,471 two-minute penalties from 2018-19 through 2025-26 that put a team on the power play with more than 2:20 left in the period. Another 5,438 came in the last 2:20 of a period. Those are priced on their own late-period line, and they’re included in the team and player totals below. For each one we compared the 2:20 after the call, even if a power-play goal ended the penalty early, with ordinary five-on-five stretches of the same length, in the same score, period and five-minute stretch of the period.
- The penalized team ends up 0.20 goals worse off than at five-on-five (95% range 0.198–0.208).
- It gives up 0.225 power-play goals per penalty and scores 0.029 short-handed goals back. Other goals in the window, after the power play ends, make up the small rest.
- In win probability, a penalty costs 3.3% of a win (3.25–3.44). That’s about one win for every 30 penalties.
3.3%
of a win, the cost of an average penalty
n = 42,471 · 95% 3.25–3.44%
30
penalties, about one win
from the average cost
8–10%
of a win, tied in the third from 10:00 to 17:40
see the period and minute chart
As a check, the method gives 0.2255 power-play goals per penalty. Counted the same way, per two-minute penalty rather than per official power-play opportunity, the league rate over the same seasons is 0.2216.
Cost of a penalty, by score
Cost to the penalized team, in % of a win, over the 2:20 after the call, against matched five-on-five stretches in the same score, period, minute and venue. Score is from the penalized team's side.
Source: HGB analysis of NHL play-by-play, regular seasons 2018-19 to 2025-26 (9,781 games). Thin lines and bands: 95% ranges.
Show the numbers
| Score (penalized team) | Cost, % of a win | 95% range | Minor penalties |
|---|---|---|---|
| Down 2+ | 1.1% | 1.0% to 1.3% | 5,589 |
| Down 1 | 3.6% | 3.4% to 3.8% | 6,987 |
| Tied | 4.4% | 4.2% to 4.6% | 14,232 |
| Up 1 | 4.2% | 4.0% to 4.5% | 8,481 |
| Up 2+ | 1.7% | 1.6% to 1.9% | 7,182 |
The score changes what a goal is worth
The power play does about the same damage in every score state: the penalized team ends up about 0.20 goals worse off each time. What changes is how much one goal moves the result.
| Penalized team | Minor penalties | Cost per penalty |
|---|---|---|
| Tied | 14,232 | 4.4% |
| Up 1 | 8,481 | 4.2% |
| Down 1 | 6,987 | 3.6% |
| Up 2 or more | 7,182 | 1.7% |
| Down 2 or more | 5,589 | 1.1% |
Up one, a power-play goal against ties the game. Down one, it turns a deficit the team was already likely to lose into a two-goal hole, so it moves the result a little less. With a lead or deficit of at least two, a single goal rarely changes who wins, and a penalty costs a quarter to two-fifths of what it costs in a tie.
Tied games: the cost climbs through the third
Tied in the first period, a penalty costs about 3.5%. Tied in the third, it costs 6.4% in the first five minutes (746 penalties), 8.2% from 10:00 to 15:00 (549) and 10.3% from 15:00 to 17:40 (259). The late samples are smaller, so their ranges are wider and overlap (see the chart).
Cost of a penalty, by period and minute
- Tied (shaded: 95% range)
- Up 1
- Down 1
Each point is a five-minute stretch of a period. The last runs 15:00–17:40, because penalties with under 2:20 left are priced on their own late-period line.
Source: HGB analysis of NHL play-by-play, regular seasons 2018-19 to 2025-26 (9,781 games). Thin lines and bands: 95% ranges.
Show the numbers
| Period, minute | Tied | Up 1 | Down 1 | Tied 95% range | Tied n |
|---|---|---|---|---|---|
| P1 0:00–5:00 | 3.4% | 3.4% | 3.5% | 3.0% to 3.7% | 2,610 |
| P1 5:00–10:00 | 3.6% | 3.3% | 3.1% | 3.2% to 4.0% | 2,206 |
| P1 10:00–15:00 | 3.5% | 3.4% | 3.2% | 3.1% to 3.9% | 1,908 |
| P1 15:00–17:40 | 3.6% | 3.4% | 4.3% | 3.0% to 4.3% | 926 |
| P2 0:00–5:00 | 4.2% | 4.1% | 3.3% | 3.6% to 4.8% | 1,315 |
| P2 5:00–10:00 | 4.1% | 3.7% | 3.4% | 3.5% to 4.8% | 1,317 |
| P2 10:00–15:00 | 4.5% | 4.6% | 3.7% | 3.8% to 5.2% | 1,079 |
| P2 15:00–17:40 | 3.9% | 3.1% | 3.9% | 2.7% to 5.0% | 619 |
| P3 0:00–5:00 | 6.4% | 6.0% | 4.3% | 5.4% to 7.4% | 746 |
| P3 5:00–10:00 | 7.8% | 5.2% | 3.4% | 6.6% to 9.0% | 698 |
| P3 10:00–15:00 | 8.2% | 5.7% | 3.2% | 6.6% to 9.8% | 549 |
| P3 15:00–17:40 | 10.3% | 5.9% | 4.9% | 7.4% to 13.3% | 259 |
Who takes the penalty: no clear difference
We tested whether a penalty costs more when the player who took it is one of his team’s regular penalty killers, since he can’t kill his own penalty. Across 40,578 penalties we added two measures for each taker, both from earlier games: his share of his team’s penalty-kill (PK) minutes, and the chances the PK allowed with him on the ice. Neither showed a clear association with what the power play produced. More than half the takers had little PK history, which limits how finely this can be measured.
Team by team, 2025-26
Here each penalty is priced at the state it was taken in: score, period and five-minute stretch. Penalties in the last 2:20 of a period use a late-period price that varies by period only. That gives every team an expected cost for the penalties it took and an expected gain for the ones it drew. These are expected values for the game state, not what happened on each power play.
- Minnesota was priced at +1.7 wins (95% range +0.7 to +2.8 wins). It drew 216 penalties and took 169.
- Toronto was at −1.3 wins (−2.4 to −0.3). It took 200 and drew 160. The Rangers were at −1.1 (−2.2 to −0.04), just short of zero.
- For the other 29 teams, the gap is too small to call either way. These ranges count only which penalties each team took and drew; the price of each game state is held fixed, which keeps them narrow. With 32 teams, one or two would land outside zero by chance alone, so this table describes the season. It doesn’t rank discipline as a skill.
| Team | Taken | Drawn | Exp. wins lost | Exp. wins gained | Net | 95% range |
|---|---|---|---|---|---|---|
| Minnesota Wild | 169 | 216 | 5.66 | 7.38 | +1.72 | +0.7 to +2.8 |
| Washington Capitals | 204 | 199 | 6.02 | 6.91 | +0.90 | −0.2 to +1.9 |
| Chicago Blackhawks | 190 | 198 | 6.43 | 7.20 | +0.77 | −0.4 to +1.9 |
| Detroit Red Wings | 182 | 216 | 6.44 | 7.01 | +0.57 | −0.5 to +1.7 |
| Anaheim Ducks | 209 | 224 | 6.90 | 7.42 | +0.52 | −0.7 to +1.7 |
| Vancouver Canucks | 194 | 197 | 5.97 | 6.49 | +0.51 | −0.3 to +1.4 |
| New York Islanders | 185 | 193 | 6.10 | 6.59 | +0.49 | −0.7 to +1.7 |
| Buffalo Sabres | 190 | 203 | 6.45 | 6.89 | +0.45 | −0.7 to +1.7 |
| Carolina Hurricanes | 194 | 199 | 6.15 | 6.58 | +0.43 | −0.6 to +1.5 |
| Vegas Golden Knights | 169 | 188 | 5.34 | 5.76 | +0.43 | −0.6 to +1.5 |
| Calgary Flames | 196 | 200 | 6.44 | 6.83 | +0.39 | −0.8 to +1.6 |
| Dallas Stars | 206 | 209 | 7.11 | 7.43 | +0.32 | −0.9 to +1.5 |
| New Jersey Devils | 168 | 172 | 5.83 | 6.08 | +0.25 | −0.8 to +1.2 |
| San Jose Sharks | 213 | 215 | 6.80 | 6.93 | +0.12 | −1.0 to +1.2 |
| Seattle Kraken | 178 | 182 | 5.74 | 5.77 | +0.03 | −1.1 to +1.1 |
| Ottawa Senators | 202 | 216 | 6.59 | 6.62 | +0.03 | −0.9 to +1.0 |
| Edmonton Oilers | 182 | 184 | 5.92 | 5.93 | +0.01 | −1.0 to +1.1 |
| Winnipeg Jets | 194 | 186 | 6.30 | 6.24 | −0.06 | −1.2 to +1.0 |
| St. Louis Blues | 172 | 176 | 5.55 | 5.48 | −0.07 | −1.0 to +0.9 |
| Boston Bruins | 228 | 210 | 7.33 | 7.20 | −0.13 | −1.5 to +1.2 |
| Columbus Blue Jackets | 185 | 161 | 5.83 | 5.57 | −0.25 | −1.4 to +1.0 |
| Montreal Canadiens | 203 | 201 | 6.93 | 6.65 | −0.27 | −1.4 to +0.9 |
| Florida Panthers | 225 | 210 | 7.38 | 7.11 | −0.28 | −1.5 to +0.9 |
| Los Angeles Kings | 188 | 189 | 6.99 | 6.68 | −0.30 | −1.6 to +1.0 |
| Utah Mammoth | 205 | 197 | 6.66 | 6.15 | −0.51 | −1.5 to +0.5 |
| Tampa Bay Lightning | 198 | 199 | 6.75 | 6.23 | −0.51 | −1.7 to +0.5 |
| Philadelphia Flyers | 198 | 195 | 6.75 | 6.05 | −0.70 | −1.8 to +0.4 |
| Pittsburgh Penguins | 204 | 186 | 6.37 | 5.62 | −0.75 | −1.7 to +0.3 |
| Colorado Avalanche | 214 | 222 | 7.38 | 6.61 | −0.77 | −1.8 to +0.3 |
| Nashville Predators | 210 | 194 | 7.23 | 6.29 | −0.94 | −2.0 to +0.1 |
| New York Rangers | 212 | 170 | 6.76 | 5.66 | −1.09 | −2.2 to −0.04 |
| Toronto Maple Leafs | 200 | 160 | 6.40 | 5.09 | −1.30 | −2.4 to −0.3 |
Expected values for the game state each penalty was taken in, not what happened on each power play. Net = gained minus lost. Ranges resample whole games and treat the per-state prices as fixed. Sorted by net; teams named above are highlighted.
Source: HGB analysis of NHL play-by-play. Expected wins, priced by score, period and time of each penalty.
Every team and season, 2018-19 to 2025-26 →
Players
Player figures are model-based expected values, with no ranges. About 9% of penalties have no recorded player who drew them, so “drawn” and net totals run slightly low.
- Connor McDavid holds four of the five best player-seasons by net since 2018-19. In 2025-26 he drew 36 penalties and took 14.
- Josh Manson (COL, 2025-26) took 30 penalties priced at 1.0 expected wins, narrowly the most of any player-season since 2018-19 (Svechnikov 2021-22 and Zadorov 2024-25 are at 0.96). Each of his penalties cost about the league average; the total comes from volume.
- Timing moves the bill. 49% of the expected cost of Ben Chiarot’s 20 penalties in 2025-26 came in the third period of one-goal or tied games. Across the league that season, the share was 26%.
| Player | Team | Season | Taken | Exp. wins lost | High-leverage share | Net |
|---|---|---|---|---|---|---|
| Josh Manson | COL | 2025-26 | 30 | 1.01 | 34% | −0.81 |
| Andrei Svechnikov | CAR | 2021-22 | 26 | 0.96 | 22% | −0.36 |
| Nikita Zadorov | BOS | 2024-25 | 32 | 0.96 | 20% | −0.61 |
| Lars Eller | WSH | 2019-20 | 21 | 0.92 | 45% | −0.48 |
| Radko Gudas | FLA | 2022-23 | 23 | 0.92 | 41% | −0.75 |
| Ian Cole | UTA | 2024-25 | 30 | 0.91 | 17% | −0.72 |
| Vincent Trocheck | NYR | 2022-23 | 22 | 0.91 | 36% | −0.53 |
| Nikita Zadorov | BOS | 2025-26 | 28 | 0.89 | 17% | −0.51 |
| Vladislav Gavrikov | CBJ | 2021-22 | 22 | 0.87 | 45% | −0.69 |
| Tanner Jeannot | NSH | 2021-22 | 20 | 0.87 | 45% | −0.60 |
Model-based expected wins, no ranges. High-leverage share: part of the expected cost that came in the third period of one-goal or tied games. Net runs slightly low: some drawn penalties have no recorded player.
Source: HGB analysis of NHL play-by-play. Expected wins, priced by score, period and time of each penalty.
| Player | Team | Season | Drawn | Taken | Net |
|---|---|---|---|---|---|
| Connor McDavid | EDM | 2023-24 | 34 | 8 | +1.04 |
| Connor McDavid | EDM | 2025-26 | 36 | 14 | +0.93 |
| Connor McDavid | EDM | 2022-23 | 35 | 13 | +0.89 |
| Jeff Skinner | BUF | 2018-19 | 30 | 9 | +0.86 |
| Connor McDavid | EDM | 2021-22 | 41 | 9 | +0.85 |
Net = expected wins gained from penalties drawn minus wins lost to penalties taken. It runs slightly low, for the same reason.
Source: HGB analysis of NHL play-by-play. Expected wins, priced by score, period and time of each penalty.