A series of tweets from @mc79hockey inspired me to write a quick post.
I’ve had several runs at modeling NHL team behavior as it pertains to OT rates, some of which contradicts the idea of these tweets. My favorite such work (on a general level) is this one, but the Sloan paper looks at the team issue more in depth. The evidence is fairly strong that some teams were smart enough to recognize that nonconference games were preferred to conference games, with respect to when teams should play OT. If that stands, its reasonable to assume that there would be some level of consistency with respect to OT rates.
That said, only 18 of a teams 82 games prior to 2013 were against nonconference foes, meaning that even if it was the case that some teams were good at playing OT in these games (see NJ’s slate here, or Pittsburgh’s here for two teams that apparently loved nonconference overtime games), comparing across years might not be able to identify a strong effect.
It’s not much different, but I figured I’d look within-season correlations of OT rates. I’d prefer the within-season numbers, as opposed to across seasons, because it keeps the players, coaches, and overall team interest in attaining points for the playoff race constant.
Splitting the current 2013-14 season in half, here are team OT rates for the first-half (x-axis) and second-half (y-axis).
It’s pretty clear from the graph above that teams have not been consistent playing for OT in 2013-14.
Looking back at every NHL season between the lockout and 2012, I did the same thing. (I excluded the 2013 season because it was shorter and teams played only their within-conference foes). For this, I used a split of January 1 to separate the “First Half” of the season from the “Second Half.”
Coincidentally, the overall correlation of team OT rates between the first half of the season and the second half was, again, -0.08.
Here’s a graph of every teams rates dating back to the 2005-06 season.
At this point, its pretty clear that teams have not been all that adept at maintaining an interest in playing OT games.
But what if we looked at nonconference games only? Here’s that same graph, using the same years and same season split, jittering the points to allow for repeats.
While its difficult to ascertain too much from the graph above, there does appear to be *some* weak relationship between how often a team played for OT in its nonconference games within a season.
One of the problems with this measurement, of course, is that with only a handful of nonconference games in each half of the season, it is likely several teams could to have an OT percentage rate of 0. As a result, the clumps of points on the axes make it difficult to distinguish if a relationship exists.
A better version of this graph would actually split each team’s nonconference season in half, and get those rates (to ensure an equal number of games on each side of the cutoff). Perhaps thats work for another day.
For what its worth, my previous work identified a subset of teams which appeared to identify the nonconference inefficiency more than others. For example, this tweet shows a plot of each team’s behavior, as judged by the log-odds of OT in nonconference games. I ran the same numbers with only these teams, and the split-season correlation of OT rates, among nonconference games, was a healthy 0.18.
In any case, this work mostly confirms the tweets from @mc79hockey. Teams, for the most part, do not appear to play for the loser point with any consistency.