Edgehalla

Prop LabGame script, momentum and scoring bias

Momentum, Game Script and Garbage Time: Which Hockey Stories Hold Up?

Hockey is full of in-game stories: momentum swings, scripted ice time, garbage-time scorers. We tested each one. Most are coin flips. Two are real, and they are not the ones people talk about most.

Verdict: Myth

By Edgehalla ResearchPublished 6 min read

Use the toolGame ScriptWho gets the minutes when a team is trailing or leading.

Momentum: real feeling, no prediction

Everyone who watches hockey has felt momentum. One team pins the other in its zone, the crowd gets loud, and it feels like a goal is coming. The question is whether that feeling predicts anything.

We built a Pressure Index to measure it. It adds up recent shots, blocked attempts, offensive-zone faceoff wins, takeaways and hits, with each event’s weight halving every 60 seconds. It is a fair picture of who is pushing play.

Then we checked it against 1,188 goals from a 200-game 2025-26 sample. The team with more pressure just before the goal scored it 51–52% of the time. A coin flip gets 50%.

Two momentum tests

The team with more pressure scored next only 51 to 52 percent of the time, and conceding a goal did not deflate a team.

Team with more pressure scored next51–52%
1,188 goals, 2025-26; coin flip = 50%
xG in the 2 min after a goal: team that conceded0.087
Expected goals created
xG in the 2 min after a goal: team that scored0.084
No “momentum swing”

The goal itself does not swing things either. In the two minutes after a goal, the team that just conceded created 0.087 expected goals and the team that scored created 0.084. If anything, the trailing side pushes a little harder, which is what score effects predict.

Pressure is still worth watching as a description of the game. It just is not a forecast. Our Pressure Index methodology explains the build and labels it “descriptive only.” This matches research on the hot hand in other sports: streaks feel stronger than they predict.

Game script and ice time

The fantasy version of momentum is game script. The idea: if a team is likely to trail, its top players will get more ice time chasing the game, so project their minutes up.

We built exactly that model. It estimated each team’s chance of winning, the share of the game it would spend leading, tied and trailing, and how each skater’s minutes change in each state. We fitted it on 2024-25 and tested it on 2025-26.

Can game script predict ice time?

Average miss on a skater’s ice time, seconds per game (lower is better)

The script-adjusted forecast missed by 130.24 seconds, slightly worse than a neutral guess at 130.19.

Neutral guess (50/50 script)
130.19 s
Script-adjusted forecast
130.24 s
Knowing the actual script (hindsight)
127.8 s
2025-26 regular season, 22,700 skater-games, model fitted on 2024-25. “Knowing the actual script” uses the real time spent leading, tied and trailing, which is not available before the game.

The forecast did not beat a neutral guess. Two things kill it. First, our pregame win probability barely predicts how long a team will trail: it explained 0.6% of it. Second, even perfect hindsight about the script only trims the miss from 130 to 127.8 seconds. Line changes, penalties and injuries swamp it.

The average expected change from game script was about 3 seconds per skater. That is too small to matter for a fantasy decision.

What is real: score-state usage

Here is the part that holds up. Coaches do use players differently when trailing and leading, and each coach’s pattern is consistent. A skater’s trailing-vs-leading ice-time tilt in odd-numbered games predicts the tilt in even-numbered games at r 0.73 (583 skaters, 2025-26).

Score effects are real at the team level too. In our margin model, a team trailing by one scores at about 1.08 times its normal rate before the last ten minutes. A team leading by one in the last ten minutes, goalie still in, scores at only 0.68 times its rate as it sits back.

So the useful question is not “will this team trail?” It is “if this team falls behind, who gets the minutes?” That is how our Game Script page frames it: scenarios, not forecasts.

Which in-game traits repeat?

Reliability of player and team traits

Six-on-five usage and trailing-vs-leading usage repeat strongly; garbage-time share and team empty-net scoring do not.

Player share of 6-on-5 time (YoY)
0.88
Usage when trailing vs leading
0.73
Team goalie-pull timing
0.30
Team empty-net goals per game
0.19
Player garbage-time share (YoY)
0.03
Usage tilt: split-half r, 583 skaters, 2025-26. Six-on-five share: year over year, 2017-18 to 2025-26. Team pull timing: split-half r (n 284 team-seasons). Team empty-net goals per game: split-half r. Garbage-time share: year over year, 226 skaters with 30+ points in both seasons.

Garbage time: a stat, not a skill

We counted a point as garbage time when the scorer’s team was up or down by three or more before the goal. About 12% of skater points come that way: 12.9% in 2023-24, 11.8% in 2024-25 and 12.1% in 2025-26.

Some players look like garbage-time specialists. But among 226 skaters with 30+ points in back-to-back seasons, a player’s share of blowout points correlated 0.03 from one year to the next. That is zero for practical purposes.

So do not draft or bench someone because their points “only come in blowouts.” Expect any player’s garbage-time share to drift back toward the league’s 12%. Our garbage-time points page shows each player’s raw share next to that regressed expectation.

Fatigue you can measure: the icing trap

One in-game effect is both real and large. When a team ices the puck, it cannot change lines. The same tired skaters must take the next faceoff in their own zone.

The icing trap

Next 5-on-5 defensive-zone faceoff: after your own icing vs normal

After its own icing, a team won 43.9 percent of the next defensive-zone draws, against 52.7 percent normally.

Defensive-zone faceoffs won
43.9%
52.7%
  • After own icing
  • Normal
200-game play-by-play sample. Expected goals allowed in the next 20 seconds: 0.022 after an icing vs 0.017 normally (+29%; 95% interval of the difference 0.002–0.008).

The tired team won only 43.9% of those draws, against 52.7% normally. It also allowed 29% more expected goals in the next 20 seconds. The effect grows with how long the defenders have been on the ice.

Tired players also take more penalties. A skater 60–90 seconds into a shift is penalized at 1.9 times the normal rate for that exposure, and 3.5 times beyond 90 seconds. The full method is in our Gas Tank and Schedule Edge methodology.

The bottom line

  • Momentum: describes the game, does not predict the next goal. Myth.
  • Game-script ice time: a forecast that did not beat neutral. Myth.
  • Garbage-time scorers: no repeatable skill. Myth.
  • Score-state usage: real and stable. Use it as “if they trail” scenarios.
  • Icing fatigue: real and large, but it happens inside games, so it is not a pregame bet.

The same lesson shows up in our line chemistry research: structure (who plays, when and with whom) repeats, while stories about streaks and special connections do not. For the other big hockey myth, see arena scorer bias on hits.

Frequently asked questions

Does momentum exist in hockey?

As a feeling and a description, yes. As a predictor, no. In 1,188 goals from 2025-26, the team with more recent pressure scored next only 51–52% of the time, barely better than a coin flip.

Does a team get a boost after scoring a goal?

Not in our data. In the two minutes after a goal, the team that conceded created 0.087 expected goals and the team that scored created 0.084. The trailing side pushes slightly harder.

Do NHL players get more ice time when their team is losing?

Some do, and it is consistent by player and coach. A skater’s trailing-vs-leading ice-time tilt repeats at r 0.73 between halves of a season. But you cannot predict before the game how long a team will trail, so use it as a scenario.

Is garbage-time scoring repeatable in fantasy hockey?

No. Among 226 skaters with 30+ points in back-to-back seasons, the share of points scored with a three-goal margin repeated at r 0.03. About 12% of all skater points are garbage time, and any player tends to drift toward that.

Does icing the puck tire out a hockey team?

Yes, measurably. After its own icing, a team cannot change and wins only 43.9% of the next defensive-zone draws, against 52.7% normally. It also allows 29% more expected goals in the next 20 seconds.

Can you predict NHL ice time from the expected game script?

Not usefully. Our script-adjusted forecast missed by 130.24 seconds per skater-game in 2025-26; a neutral guess missed by 130.19. Even knowing the real script afterward only cut the miss to 127.8 seconds.

More in Game script, momentum and scoring bias

All Prop Lab research

Sources and method

  1. NHL stats (NHL.com)
  2. NHL API reference (community documentation)
  3. Hot hand (research overview)
  4. Split-half reliability and the Spearman–Brown formula
  5. Kaplan–Meier estimator

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