PULL THE
GOALIE
PLAY

WHEN TO PULL THE GOALIE

Down one at even strength, pull him with about 2:30 left. Down two with 5:00 or less left, pull him now. For any other game, set it below.

SET THE GAME

SCORE

In the third, 0:15 to 20:00. Type it like 2:30 or 230.

STRENGTH

THE MODEL'S CALL

DOWN 1 · 3:00 LEFT · EVEN

NOT YET

PULL HIM AT 2:27

Any time from 2:45 to 2:10 left is as good, to 0.01%.

WIN CHANCE

PULL AT 2:278.39%
PULL NOW8.36%
NEVER PULL4.56%

Pull at 2:27 and you tie it before the horn about 17% of the time, against 9.1% never pulling; they score into the empty net about 44% of the time, but losing by two costs no more than losing by one.

WHAT NHL TEAMS DID

In close 2025-26 games, NHL teams down one pulled the goalie with 1:54 left on average (355 pulls). See every team in the Goalie Pull Report.

SHARE THIS ANSWER

EMBED IT

A plain link for your blog or forum post. It opens this answer.

<a href="https://pullthegoalie.app/when-to-pull?down=1&amp;left=3:00">When to pull the goalie: Pull the Goalie's calculator</a>

QUICK ANSWERS

At even strength, from each clock: when the model pulls the goalie, and how often that plan wins.

LEFTDOWN ONEDOWN TWO
1:00NOW
WINS 4.05%
NOW
WINS 0.16%
1:30NOW
WINS 5.57%
NOW
WINS 0.31%
2:00NOW
WINS 6.83%
NOW
WINS 0.49%
2:30NOW
WINS 7.64%
NOW
WINS 0.71%
3:00AT 2:27
WINS 8.39%
NOW
WINS 0.92%
4:00AT 2:29
WINS 9.84%
NOW
WINS 1.32%
5:00AT 2:30
WINS 11.19%
NOW
WINS 1.69%
NHLAT 1:54
355 PULLS
AT 3:18
130 PULLS

NHL: the average time left when teams down one and down two really pulled, in close 2025-26 games at any strength (the Goalie Pull Report).

HOW IT DECIDES

This is the model behind every grade in Pull the Goalie, fit on 3,936 NHL games from three seasons. For every plan, from pulling now to never pulling, a second apart, it plays out the rest of regulation second by second at the measured scoring rates and adds up the chance of winning. The call is the plan that wins most. A tie counts as half a win.

Pulling the goalie gives up empty-net goals far faster than it scores, and it is still worth it: down one, a two-goal loss costs no more than a one-goal loss, so the only number that matters is the chance of tying it before the horn.

What it leaves out. Every team is an average team, the puck is in play (no faceoff), and the plan holds whatever happens next: if the score changes, set it again. A penalty is a two-minute minor that runs its whole time, even if the power play scores. Some special-teams rates are too rare to measure (every short-handed rate with the net empty, and a few others), so the model derives them from the five-on-five or empty-net rate and the measured effect of the power play or the kill. Down three, the game counts the night as lost.

The long version is when to pull the goalie on the strategy page, and how the model works, with the charts.

PLAY IT