The Goalie Pull, Priced
Every initial deficit goalie pull across three NHL seasons (2023-24 through 2025-26), graded the way a moneyline settles: did the team that emptied its net go on to win the game, overtime and shootout included. The last column is the moneyline that makes each spot break even.
Down one goal
The only spot with a real comeback rate. Between one and three minutes it is essentially flat, 8.37% to 8.80%, so waiting out an extra minute of that window costs nothing. Inside the final minute it collapses to 0.53%, 1 from 187 pulls, which is not a worse price so much as a different bet. The 3:01-10:00 bucket reads higher at 6.12%, but on 49 pulls its 95% range runs 2.1% to 16.5%, so treat it as a hint rather than a better spot.
| Spot | Clock at pull | Pulls | Wins | Win rate | 95% range | Breakeven price |
|---|---|---|---|---|---|---|
| Down 1 | 3:01-10:00 | 49 | 3 | 6.12% | 2.1-16.5% | +1533 |
| Down 1 | 2:01-3:00 | 432 | 38 | 8.80% | 6.5-11.8% | +1037 |
| Down 1 | 1:01-2:00 | 753 | 63 | 8.37% | 6.6-10.6% | +1095 |
| Down 1 | 0:00-1:00 | 187 | 1 | 0.53% | 0.1-3.0% | +18600 |
Down two or more
Teams pull earlier here, and it almost never works. These rows exist to show how far the price has to move before the bet is even arguable.
| Spot | Clock at pull | Pulls | Wins | Win rate | 95% range | Breakeven price |
|---|---|---|---|---|---|---|
| Down 2 | 3:01-10:00 | 457 | 13 | 2.84% | 1.7-4.8% | +3415 |
| Down 2 | 2:01-3:00 | 355 | 3 | 0.85% | 0.3-2.5% | +11733 |
| Down 2 | 1:01-2:00 | 138 | 1 | 0.72% | 0.1-4.0% | +13700 |
| Down 2 | 0:00-1:00 | 28 | 0 | 0.00% | 0.0-12.1% | no wins |
| Down 3+ | 3:01-10:00 | 256 | 1 | 0.39% | 0.1-2.2% | +25500 |
| Down 3+ | 2:01-3:00 | 35 | 0 | 0.00% | 0.0-9.9% | no wins |
| Down 3+ | 1:01-2:00 | 12 | 0 | 0.00% | 0.0-24.3% | no wins |
| Down 3+ | 0:00-1:00 | 1 | 0 | 0.00% | 0.0-79.3% | no wins |
Season by season, down one
Worth reading before trusting the headline number: the down-one rate swings by season on samples of 450 to 492 pulls, which is the sample size talking as much as anything about the teams.
Method, and what this does not say
- Pulls come from the NHL play-by-play feed, which flags an empty net in its per-play situation code. A pull counts once per game: the first empty-net episode for the trailing team in the third period that begins with 10:00 or less left. Delayed-penalty empty nets are excluded.
- That 10:00 cap is a hard edge, not a bucket boundary. Anything earlier is outside the study entirely, which is why the widest row reads 3:01-10:00 and there is no row above it. Only 139 of these pulls came before the 5:00 mark at all, and just 1 of those were down one goal, so the top bucket is in practice a 3:00-to-5:00 bucket.
- The feed has no discrete "goalie pulled" event, so the clock is taken from the first play after the net shows empty. Every time in the table therefore runs slightly late, by seconds rather than minutes.
- A win means the pulling team won the game. Tying it and losing in overtime is a losing moneyline, and it is counted as one.
- These are breakeven prices, not results. They say what a sportsbook had to pay you for the bet to clear, not what it actually paid. What the board actually showed at each of these moments is a separate read and is not on this page yet, so nothing here is a claim that any of these spots was profitable.
- Small cells are noisy. The 95% range column is a Wilson interval; where it is wide, treat the win rate as a hint rather than a number.
- Which teams pull earliest, and by how much, lives on the NHL goalie pull tracker.