Paper:  Pulling the Goalie

Introduction: Cliff Asness & Aaron Brown (of AQR Capital Management) wrote an interesting paper on the concept of “pulling the goalie” in Hockey. What starts as a simple thought experiment turns out to have implications far beyond the realm of sports.

Commentary:

I was exposed to this paper as well as SSRN (Social Science Research Network) all thanks to Malcolm Gladwell and his Revisionist History Podcast. I believe this paper is the perfect launching point for this blog because the thesis is the epitome of what I love about empirical data and how much of it runs contrary to many ideas and preconceived notions that we as a larger society take as fact. I find it incredibly interesting to find this specific points and bringing them to light.

Now I do not pretend to be an expert on hockey or any sport for that matter, but this example of a common thought that doesn’t make much sense when put under the microscope. It is a common strategy for a hockey coach to pull his goalie with around 1 minute left in a game in which they are losing in the hopes that with one less defender and one more offensive player that there is a greater probability of a score. This is widespread and doesn’t vary much from team to team or levels of play. Now, why is this? Well, people have used this 1-minute point for a long time, so it is believed to the optimal time for the switch to be made.

What Assness and Brown uncovered is that this tradition is far from optimal. In fact, they found that the best time to pull the goalie (when down by a single score) is actually with 5:40 left in the game. This optimal time pull will give a team “gains [on] average of 0.02 more points per game.” While this seems like a small, possibly insignificant number on aggregate this is “worth 1.76 points in an 82-game season, over a team that never pulls the goalie.” Furthermore, when a team is down by 2 points, the goalie pull-time increases to more than 10 minutes – this is unheard of in traditional hockey play.

So the question I ask is: “Why?”. Why would coaches not pull their goalies earlier is there is a significant benefit to doing so? I concur with the conclusion stated by Assness and Brown that coaches don’t really have the incentive to do so. This is because coaches would rather lose a game by 1 goal than risk losing a game by 2 or more. Fans and hockey pundits would call into question the coaches ability to manage their team and their decision-making ability. Coaches are not rewarded for winning games, but for being perceived as good coaches. Coupled with this is the fact that “winning ugly is undervalued versus losing elegantly, and losing ugly can be career suicide” for the coach.

An interesting disparity between conventional thought and empirical efficiency.