On May 25th, I watched a Finland team with 0 NHL players defeat a loaded Russian squad in the 2019 IIHF World Hockey Championship Semi-final. Russia is a team with many NHL players and superstars like Alex Ovechkin, Evgeni Malkin and Ilya Kovalchuk. Then there was Finland; a team without a single NHL player. The game was scoreless until the 9:42 mark of the third period when Marko Anttila scored to put Finland up 1-0. Russia then waited until there was one minute and 58 seconds left before they pulled goalie Andrei Vasilevski. Russia then peppered Finland’s defense with shots for the remaining two minutes, but Finland was just able to escape without allowing a goal for a 1-0 final.
It seemed crazy to me that Russia had lost and that Finland had somehow kept the puck out of the net. My thought after watching was that if you pull your goalie and don’t score or even allow a goal, then you waited too long to pull your goalie. How many times do you think, “man if we only had one more minute”. My thought is why don’t you plan ahead and give yourself another minute then, or more. By pulling your goalie you gain an extra-attacker and are attacking 6-on-5 against the defense. You control the puck much longer and have many more shots on goal. Yes, you leave your net open, but you’re losing so you need to take the risk. I decided to look further into when teams should pull their goalies. And I’ll tell you what I learned is NHL teams are doing it wrong and national teams are doing it wrong. Everyone is doing it wrong.
I will base a lot of the numbers in this post off of research done by Clifford Asness (U of Chicago grad, billionaire hedge fund manager and co-founder of AQR Capital Management) and Aaron Brown (Harvard and U of Chicago grad and former Chief Risk Manager at AQR Capital Management). In the NHL, there is around a 0.65% chance a team scores in each 10 second interval of a game. When you pull your goalie your chance increases to 1.97% while also increasing the chance your opponent scores to 4.30%. So obviously, pulling your goalie makes it much more likely that your opponent scores than you do. But that’s not what we should be looking at. We should look not at goals as a measurement of value here, but at what a team stands to gain by scoring. A goal for the trailing team means substantially more than a goal to the leading team. This is where risk management comes into play; the losing team needs to risk more because they have more to gain.
Your team is down, so you’re already in a bad position. You have everything to gain by scoring, it brings your team right back into it and brings your probability of winning back to 50%. The graph below shows a team’s chances of winning when trailing in an NHL game.
What Brown and Asness found in their research is the optimal time to pull a goalie based on the number of goals down by and time left in the game. They built a model, which is too complicated for me to explain here (but I will link to their paper at the bottom). The results of their model can be seen below (Figure from Asness and Brown paper).
The optimal time to pull your goalie when down one goal is actually with six minutes and 10 seconds left and if you’re down two goals, it’s with 13 minutes left. If you score at 13 minutes, you would then replace your goalie until the 6:10 mark. Pulling your goalie is a risk, but only pulling your goalie with two minutes left is not the most effective risk. The best risk for your team’s chance to win is to pull the goalie with six minutes left.
If the best chance to give your team a win is pulling your goalie with six minutes left, then why don’t teams do it? Because that’s not tradition. You wait until there’s one or two minutes left because that’s how hockey has always been played. The worst reason you can give for anything; because that’s how it’s always been. The next reason is that there is a high chance you do in fact give up empty-netters and lose some ugly games. The fans can turn on you and may feel you robbed them of a more competitive finish. You will have a larger margin of defeat when you lose, but so what- that doesn’t cost you any points in the standings. Fans are going to be pissed when it doesn’t work, and most of the time it won’t work. However, it will allow you to tie more games than if you wait until there’s two minutes left.
Failing to pull their goalie hurt Russia last week and it hurt them in one of the most famous games of all-time. In the 1980 Olympics between the Soviet Union and the United States, the USA led 4-3 with ten minutes to play. But Russian coach Victor Tikhonov never pulled his goalie. Players on the Russian team said he didn’t believe in it and they never even practiced 6-on-5. The underdog Americans were able to hold off a furious 5-on-5 Soviet attack in one of the greatest sports upsets in history. But if Tikhonov had pulled the goalie, we may not have ever had the Miracle on Ice we had in 1980.
It’s not just Russia though, NHL coaches don’t pull their goalies near enough either. If coaches want to give their team the best chance to win, they will follow this model, but it will take a coach who doesn’t care about the backlash and the fans who will call him an idiot when they give up an empty netter with six minutes left. It’ll be a coach whose only focus is winning.
Clifford Asness and Aaron Brown’s Paper: SSRN-id3132563