clock menu more-arrow no yes mobile

Filed under:

The risk/reward of pulling the goalie

Patrick Roy had a knack for pulling his goalie early in games where his team trailed. Was he on something, or onto something?

NHL: Colorado Avalanche at Arizona Coyotes
“Get over here Varly.” - Patrick Roy, probably
Matt Kartozian-USA TODAY Sports

The art of pulling the goalie is one that takes a fair amount of awareness with how the game is going. In most games, a goalie would be called to the bench whenever a penalty is pending against the opposing team. In rare instances, you might find a coach elect to yank the netminder with an offensive zone draw late in any given period. This is extremely rare however, and shouldn’t be thought of as a regular occurrence.

One of the great spectacles in sports however is when your team is playing with a deficit on the scoreboard, and late in the 3rd period elect to pull the goalie and replace him with an extra attacking skater on the ice in the hopes it will lead to a game-tying goal. How often does that work though?

A man named Andrew Thomas studied the timing of goals being scored in the NHL over 4 seasons between the 2002-03 season through the 2006-07 and published his results in the Journal of Quantitative Analysis in Sports. The study is an interesting one which focuses more on the value of goals scored at any given moment in an NHL game. In it though, there is a decent enough chance that pulling your goalie to make it worth the risk. As the analysis reports:

“During this period, the likelihood of one team scoring is markedly increased, since the trailing team has one extra offensive play and their net is un-guarded. The situation is quite different from an ordinary game situation, and so is take int account separately: when a goaltender is pulled and indication is made that the next interval is being played with an empty net. In the data being considered, a goaltender was pulled in 64% of games played. In 34% of those game a goal was scored with one net empty, and of those goals 30% were scored by the team that pulled their goaltender”

The caveat with the data Andrew was analyzing is that there is no indication from his sample size as to why the goalie was being pulled. Whether it was for an impending penalty or it was a last ditch effort to tie the game was not clearly stated by the data being analyzed. What Andrew did to account for this was to limit his research on this specific matter to the final 2 12 minutes of the games in his sample where the chances of a goalie being pulled were increased for gaining an extra attacker over a pending penalty.

The data however suggests that a goal is likely to happen 1 out of every 3 times a goalie is pulled for the extra attacker, either by the trailing team or a goal into an empty net. The risk of pulling the goalie is not so great that you might think twice before emptying the cage. It goes on to state that 3 out of 10 goals scored with the extra attacker. Essentially, if there is 100 instances of pulling a goalie, there would be 34 goals scored, about 10 of which would be potential game-tying goals scored by the team with the empty cage.

All this really points to is that empty net goals are rare, for either side of the ice. The risk born by the trailing team is not terribly high. You’re already losing anyway, so running this risk is minimal. The data also seems to point to former Colorado Avalanche coach Patrick Roy not being quite as crazy as we might have made him out to be. He seemed to have no qualms about pulling his goalie out of games much earlier than most believe you should to gain an advantage and a tying goal. While the risk is there, a loss is a loss no matter what the score, but a potential win or at the very least a potential loser point in overtime can justify the risk inherent in pulling the goalie.

In today’s NHL, points are at a premium. There is no set number of points awarded every season, but instead a range of points that can be awarded. With 30 teams in the league currently, and if every game was won in regulation in a season, the NHL would award 2460 points. If every game went to overtime, there would be 3690 points awarded across the board, a 1230 point difference. That flawed system in a league ripe with parody where even 1 point can mean the difference between playoff glory and an early offseason, reduces the burden of pulling your goalie where the reward is not only at least a point in the standings, but a ticket to the playoffs as well.