Coming in Hot: More on Momentum and the Playoffs

This week we looked at whether there was anything behind the common claim that teams need to go into the playoffs playing well. There are actually several ways one could examine this problem, some of which unfortunately suffered from a lack of data. Ideally, I think we would have liked to look at a team’s “fundamentals” over the last several games of a season to get an idea of how they were playing – specifically, we would have liked to consider their Corsi or Fenwick Close. But, we could only get this data going back to 2008, which meant only 6 years of playoff series. Unfortunately, it just wasn’t enough.

So, instead we looked the team’s record over the last several games. As mentioned in The Star column, we considered the probability of winning a playoff series as a function of the point differential between the two teams as well as the point differential over the last stretch of games. The point differential attempts to capture the relative quality difference in the teams (there is a favourite and an underdog, and the extent to which one team is favoured over the other varies from series to series). We could have considered looking at each team’s points separately, but the inclusion of the Bettman loser point halfway through the sample period meant that 90 points pre-lockout does not mean the same team quality as 90 points post-lockout. Looking at the differential, however, should go a long way to eliminating this kind of “measurement error”.

For momentum, we considered the point differential between the two teams over the last 5 games as well as the last 10 games. It wasn’t obvious to us how long a team had to be playing well in order for it to be considered as “having momentum”, so we figured we should consider at least a couple of different possibilities.

Finally, we thought we should try to get an idea of how long any “momentum” might last, so we used a sample of all series as well as a sample of just first round series.

As we did in the hitting piece, we used a probit regression model to figure out how these variables translated into probabilities of winning. For a discussion of this model, go here.

I thought our results were pretty interesting. When we looked at all playoff series (a sample size of 225), we didn’t get a statistically significant effect for either the 5 game or 10 game measures of momentum. However, when we looked at just the first round (120 series), the 5 game measure was not statistically significant, but the 10 game measure was. The regression results can be seen here.

My take on this is that there is something to the idea that teams want to be playing their best hockey as they get to the playoffs. The reason that no effect shows up when we look at all series, however, is that how well a team is playing can change rather quickly. It could happen even during the course of the first round. Thus, the further a team goes into the playoffs, the less correlated the team’s record at the end of the regular season is with how they are currently playing.

As for the difference between the 5-game measure and the 10-game measure, my guess is that 5 games is just too short a period to have much of a correlation between how a team is playing and its record. Anyone following the Leafs this year could see that a team can be playing pretty terrible hockey and still post a good record over a 5 game stretch. Yes, they also posted good records over 10 game stretches, but this becomes less likely, so there’s a stronger correlation between the team’s record and how they’re playing.

Also, teams that have locked up their playoff seeding with 2 or 3 games left have the luxury of resting players, which will further skew the results.

1 Comment

  1. brian smith's Gravatar brian smith
    June 10, 2014    

    so you so called statistical genius;s wheres your plain awful goalies NOW. looks like all of you should keep your day time job and let the game be decided on the ice.Another so called experts taking up print in the sports page when really stats dont mean a thing brian smith

Leave a Reply to brian smith Cancel reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>