Empirical evidence of short term 'momentum' in the NHL?

Comebacks vs. blown leads in overtime games.

Posted by Mike Apted on Saturday, January 16, 2016

My exploration of SDQL continues, and as before I’ve just been working to pick up the basics of the query language. ‘Momentum’ is an oft debated concept in sports. The question this time is:

Do teams that blow leads typically also lose the game in overtime?

In other words, does a team charging through a comeback carry that ‘momentum’ into overtime? Using KillerSports.com as my source (the data goes back to 2006) and filtering for regular season only I decided to look at teams winning percentages in overtime when holding a given lead after one or two periods of play.

To start I looked at winning percentages of teams, in games that reached overtime, with a lead after one period:

Goal Diff After 1st Period 1 2 3 4
OT Winning Percentage 50.1% 51.5% 48.7% 50.0%

Some slight variance but generally it appears that the score after the first period doesn’t impact the “crap shoot” that is overtime/shootouts in the NHL. So then I looked at the same data for teams with a lead after two periods of play:

Goal Diff After 2nd Period 1 2 3 4
OT Winning Percentage 50.7% 46.1% 37.5% 40.0%

In this case it seems like a teams winning percentage in overtime is significanlty lower given the size of the lead they surrender in the third period. The 4 goal blown lead after two periods only occurs 5 times (3 of 5 games) so it’s conceivable that number could be different with a better sample.

So in summary it looks like overcoming even a large first period deficit to reach overtime doesn’t affect the winning percentages, however, a team coming back in the third period seems to carry a proportional amount of ‘momentum’ into the overtime/shootout frame. Perhaps given the lack of a 20 minute intermission helps, maybe the comfort of knowing they at least maintained the single point (in today’s scoring system) affects the team that surrendered the lead?

Examples of the basic SDQL used to generate these results (updating the goal differential) was:

(period scores[0]) - (o:period scores[0]) = 4 and playoffs = 0 and overtime


(period scores[0] + period scores[1]) - (o:period scores[0] + o:period scores[1]) = 1 and playoffs = 0 and overtime

Again, As I get more comfortable with SDQL I am hoping I can find some more insightful questions to answer, but this continues the “get your feet wet” phase.