When a team enters an intermission on the power play, they don't just score during the man advantage — the offensive advantage continues well into regular 5-on-5 play after the penalty expires.
A carryover power play happens when a team is still on the power play when the buzzer sounds to end a period. Both teams go to the locker room and the power play resumes at the start of the next period — from a centre-ice faceoff, not the offensive zone.
The question: does starting a period on the power play actually help? Earlier research said no — or even that it hurt. Our five-season study of 6,365 NHL games finds the opposite.
Carryover teams score more often during the active power play, but that's only part of the story. They continue to outscore opponents even after the penalty expires and teams return to even strength. That lingering advantage — the "afterburn" — is the most important and surprising finding in this study.
Carryover teams score 4.3 percentage points more often in the five minutes after a period restart — and most of that advantage happens after the penalty has already expired, suggesting that carryover situations create a lasting offensive ripple that outlasts the man advantage itself.
Two prior studies shaped the conventional wisdom. Both were limited by sample size and, critically, by how they defined the comparison group. Our study fixes that — and the picture changes substantially.
The most-cited prior study (Seattle Kraken / Sportlogiq) measured whether carryover teams scored on the active power play itself — and compared it against mid-period power plays. The problem: mid-period power plays start from an offensive-zone faceoff, which is far more dangerous than a centre-ice restart. Comparing those two situations is like comparing a penalty shot to a regular power play and concluding power plays are weak.
Our study asks the right question instead: compared to teams that start a period without a power play, do carryover teams score more in the following few minutes? The answer is clearly yes — and the mechanism is mostly not what people assumed.
Here is the most important table in this study. The active power play scoring rate barely moves between 120 seconds and 300 seconds — from 9.9% to 10.5% — because 88% of all carryover penalties are 2-minute minors that expire right at the 120-second mark. But the broad scoring rate keeps climbing, reaching 25.2% by five minutes. That continued scoring is happening at even strength, after the penalty is gone.
| Time After Restart | Active PP Scoring | All Carryover-Team Scoring | The Afterburn Gap | What's Happening |
|---|---|---|---|---|
| 0–30 seconds | 2.2% | 2.5% | +0.3pp | Teams still setting up from centre ice |
| 0–60 seconds | 6.2% | 7.5% | +1.3pp | PP generating most value; small spill |
| 0–120 seconds | 9.9% | 14.5% | +4.6pp | 2-min minors expiring — post-PP momentum starting |
| 0–300 seconds | 10.5% | 25.2% | +14.7pp | Penalty long gone — momentum effect dominates entirely |
If carryover teams only benefited while the power play was active, broad scoring at 300s would be close to the active PP rate (~10.5%). Instead it reaches 25.2% — a 14.7pp gap that is almost entirely occurring in 5-on-5 hockey. Carryover situations create a ripple effect that survives the whistle ending the power play.
Every result below is statistically significant and survived correction for multiple comparisons (Benjamini-Hochberg method). An odds ratio above 1.0 means carryover teams are more likely to score. The control group is period restarts with no active power play.
| Window | Situations | Goals | Carryover % | 95% Confidence Range | Control % | Advantage | Odds Ratio | Significant |
|---|---|---|---|---|---|---|---|---|
| 30 sec | 2,898 | 71 | 2.5% | [1.9% – 3.1%] | 1.5% | +0.9pp | 1.65 | Yes ✱ |
| 60 sec | 2,898 | 217 | 7.5% | [6.6% – 8.5%] | 3.6% | +3.9pp | 2.19 | Yes ✱ |
| 120 sec | 2,898 | 420 | 14.5% | [13.3% – 15.8%] | 8.3% | +6.2pp | 1.86 | Yes ✱ |
| 300 sec | 2,898 | 731 | 25.2% | [23.7% – 26.8%] | 20.9% | +4.3pp | 1.27 | Yes ✱ |
After adjusting for home ice, score differential, remaining penalty time, and season — the carryover advantage at 300 seconds has an odds ratio of 1.40 [1.22–1.60]. The advantage is not a product of game context; it's a real effect.
The carryover broad advantage over controls appears every year — from +2.6pp in 2020–21 to +5.6pp in 2023–24. The active PP rate has also trended upward season-over-season (9.1% to 11.7%), suggesting improving PP efficiency on restarts in recent years.
| Game Type | Carryover Score % | Control % | Active PP % |
|---|---|---|---|
| Regular Season | 14.8% | 8.4% | 10.1% |
| Playoffs | 11.5% | 7.1% | 8.0% |
The carryover advantage shrinks in playoff games — 11.5% vs. 14.8% at 120 seconds. Playoff penalty kill units are better organised and more prepared for period-restart scenarios. The effect doesn't disappear, but structured preparation reduces it — which is itself an actionable finding.
The most important contextual fact: 88% of all carryover situations involve a standard 2-minute minor penalty. These expire exactly at the 120-second mark of our time windows. So when scoring keeps climbing from 120s to 300s — with the active PP rate barely moving — all of that additional scoring is happening at 5-on-5 strength. The afterburn is real and it is provably not just "more power play time."
| Penalty Length | Count | Share | Expires Within 300s? |
|---|---|---|---|
| 2 minutes (minor) | 2,556 | 88.2% | Yes — at exactly 120 seconds |
| 4 minutes (double minor) | 58 | 2.0% | Yes — at 240 seconds |
| 5 minutes (major) | 191 | 6.6% | Right at the 300s boundary |
| 10 minutes (misconduct) | 83 | 2.9% | No — extends beyond window |
| 15 minutes | 10 | 0.3% | No — well beyond window |
The play-by-play data tells us that the afterburn exists but can't fully explain why. Four mechanisms likely work together:
Power plays often end with the puck in the offensive zone and defenders pinned back. When the penalty expires, teams don't automatically reset — the attacking team can frequently maintain possession and pressure through the transition into 5-on-5.
PK players who begin the period defending immediately — without a warmup shift — start at a physical disadvantage relative to PP players, who can select their freshest personnel for the restart. That fatigue doesn't vanish when the penalty clock hits zero.
The carryover team controls which players are on ice for the restart and can keep their top offensive line out after the expiry. The defending team may have committed defensive personnel to the PK that are now caught in the wrong position at even strength.
PP pressure generates secondary chaos: failed clears produce icings, icings force defensive-zone faceoffs, and sustained pressure can draw additional penalties. These downstream events keep the advantage alive well after the original power play expires.
Earlier research on period-start ice conditions found that fresh ice does not help attackers at even strength. In a study of 3,936 games and 16,250 goals, goal share was consistently below time share on fresh ice — if anything, goalies perform slightly better on clean surfaces.
This is consistent with our findings, not contradicted by them. The carryover advantage does not come from the ice surface — it comes from the tactical and physical state of both teams entering the period. The fresh-ice effect is a wash. The carryover advantage is real despite the ice, not because of it.
Drawing a penalty late in a period isn't just about the immediate power play. It generates a broader offensive advantage that outlasts the man advantage. Teams should treat these as higher-leverage situations than an equivalent mid-period penalty.
Most coaching attention goes to the power play setup itself. The bigger opportunity may be planning what happens at the 120-second mark — when the penalty expires. Scripted forecheck pressure and pre-planned line deployment in that window could be where the real advantage is captured.
For the PK team, the data says: when the penalty expires, prioritise getting the puck deep and forcing a line change. Don't let the carryover team maintain possession through the transition. Playoff PK units do this better — the gap shrinks significantly in the playoffs.
Team power-play percentages blend very different situations together. Carryover power plays have a different starting position, different tactical context, and substantially different downstream value. They deserve their own column in any serious team report.
Because carryover periods restart at centre ice rather than the offensive zone, the opening faceoff is the first critical moment. Winning that draw cleanly and executing a set entry is where the active PP window is won or lost — this is a directly coachable skill.
A penalty drawn with 45 seconds left in a period is not the same as one drawn with 3 minutes left. The carryover situation adds value a mid-period PP does not. Teams should be more aggressive about drawing calls in the final minute of a period.
Reported carryover active-PP conversion at 10.8% (2020–21) and 17.0% (partial 2021–22), both below the league-wide PP rate of ~20–21%. This led to the conclusion that carryover situations underperform.
The flaw: it compared carryover power plays (which start at centre ice) against mid-period power plays (which start from the defensive zone — far more dangerous). Our study uses the right control: period restarts with no carryover. That change in comparison group reverses the conclusion.
Analysed 3,936 games and 16,250 even-strength goals across three seasons. Found that period-start fresh ice does not increase scoring — goal share runs below time share on fresh ice, and goalies may actually benefit more from clean surfaces.
This is fully consistent with our results. The carryover advantage is not an ice-quality effect. Our finding and this finding coexist without contradiction — they are describing different mechanisms.
| Study | Sample Size | Method | Finding | Key Limitation | Status vs. This Study |
|---|---|---|---|---|---|
| Kraken / Sportlogiq | 2020–22 partial seasons | Active PP conversion vs. mid-period PPs | Carryover PP converts below league avg | Wrong control group — compares centre-ice restarts to offensive-zone starts | Partially superseded — active-PP rate is real but the framing is misleading |
| Fresh Ice (Sportlogiq) | 3,936 games, 3 seasons | Time share vs. goal share by ice condition | Fresh ice does not help attackers at even strength | Even-strength only; no PP or carryover context | Confirmed and compatible — different mechanism, same conclusion |
| This Study (NEA, 2024–25) | 6,365 games, 5 seasons, 2,898 carryovers | Broad + active outcomes, proper non-carryover control, logistic regression with covariates | Carryover teams score +4.3pp more in 5-min window; downstream 5-on-5 effect is the dominant mechanism | Observational — mechanisms inferred from data, not experimentally isolated | Original contribution |
Play-by-play data was pulled from the NHL GameCenter API (api-web.nhle.com/v1/gamecenter/{id}/play-by-play) for all regular season and playoff games across five seasons: 2019–20 through 2023–24 — 6,365 games, 90,248 exposure rows. Each game's full event log was cached as gzipped JSON and parsed in Python using pandas.
The penalty timeline was rebuilt event-by-event for every game. Each penalty's start and end time was tracked, with 2-minute minors shortened at the exact game-second a power play goal was scored (as the rulebook requires). This produced a second-level manpower state record used to identify which team — if any — was on the power play at the final second of each period.
Treatment: any period transition (P1→P2 or P2→P3) where one team was on the power play at the horn — 2,898 situations. Control: period transitions where neither team held a man advantage — 19,664 situations. This is the critical design choice that distinguishes our study from prior work, which compared carryover PPs against mid-period power plays instead.
Two outcome variables were tracked at four time windows (30s, 60s, 120s, 300s) after the next period's opening faceoff:
Broad: any goal scored by the carryover team within the window, regardless of manpower state. This captures both active-PP goals and post-expiry 5-on-5 goals.
Active PP: a goal scored exclusively while the carryover power play was still in force. This replicates the definition used by prior research.
Fisher exact test was used for each window (preferred over chi-square given small carryover cell counts in some splits). Benjamini-Hochberg FDR correction was applied across the four-window family to control false discovery rate at α = 0.05. All four windows survived correction.
Logistic regression (statsmodels) was fit for each window with covariates: home/away indicator, score differential for the PP team at period end, remaining penalty seconds at the horn, and season fixed effects. The carryover coefficient and its odds ratio are the adjusted estimates reported throughout.
This is an observational study — causal mechanisms are inferred from patterns in the data, not experimentally isolated. Empty-net situations are included in the primary specification (conservative). The active PP logistic regression produces degenerate estimates due to complete separation (the control group has zero active-PP goals by definition); Fisher exact and descriptive rates are the appropriate inference tools for that comparison only.
| Parameter | Specification |
|---|---|
| Data source | NHL GameCenter PBP API + NHL Stats REST (schedule enumeration) |
| Seasons | 2019–20, 2020–21, 2021–22, 2022–23, 2023–24 |
| Game types | Regular season (game_type=2) and playoffs (game_type=3), analysed separately and combined |
| Transitions | P1→P2 and P2→P3 only (OT excluded) |
| Treatment (n) | 2,898 carryover PP situations across 6,365 games |
| Control (n) | 19,664 non-carryover period restarts |
| Time windows | 0–30s, 0–60s, 0–120s, 0–300s from restart faceoff |
| Primary test | Fisher exact, two-sided, BH-corrected (α = 0.05) |
| Adjusted model | Logistic regression: goal ~ carryover + home_pp + score_diff_pp + rem_pen_sec + C(season) |
| Penalty expiry | 2-min minors shortened on PP goal per rulebook; majors run full duration |
| Tools | Python 3.10 · pandas · scipy · statsmodels · Oracle Cloud VM (Ampere A1, Ubuntu) |