Sunday, March 22, 2009

Playing to the Score

This is a first look at the effect of playing to the score. I will be doing some more work on this in the days to come but thought I would throw some preliminary findings out before doing too much more so that I can add in some improvements if others have some suggestions. So feel free!

As the title suggests, the topic at hand is playing to the score. When I began I decided, rightly or wrongly, that a team is likely to play to the score in close games in the third period but is unlikely to play to the score in the first period. As such I tracked all of the games that have been played in the 2008-2009 NHL season in which the goal differential between the two teams was no greater than two and where no goals were scored in the third period. I then compared the shot totals in the first period of these games with the shot totals in the third period of these games. One of the limitations of this study is that it limits the sample size but I couldn't find an easy way to strip out the different situations easily in games where the score changed frequently. I would also like to have used Corsi numbers instead of shots but couldn't find a place that breaks these numbers down by period. Anyway, we are left with 142 games to this point in the season, 58 in which the teams entered the third period tied and 84 in which one team held a one or two goal lead. The results for the first period do not reflect the game state in that period, so the column that says "First Period Tied" means that these are the shots that were taken in the first period of games that were tied in the third period. Similarly "First Period Trail" are the shots that were taken in the first period by the team that trailed for the entire third period, and so on. These are the shot totals for those games:


In order to get the lead teams did not need to significantly outshoot their opposition, at least in the first period. It may be, of course, that teams got their leads in the second when they had more success outshooting. Nonetheless, this is one piece of evidence that may suggest that outshooting does not necessarily result in outscoring.

Once teams have the lead in the third period they have a tendency to sit back. This is what most observers of the game would intuit and so I think this passes the smell test. It also, to a degree, answers the question of how much: teams that are trailing increase their shot total by 11.3%, teams that are leading decrease their shot total by 21.1% and the overall shots in games where one team has the lead decrease by 5.2%. Given that these are instances in which neither team scored a goal it is quite possible that shot quality is going down for the teams increasing their shot output even as the number of shots is going up.
In tie games there is far more shot attrition with neither team being forced to push the action: there are 21.3% fewer shots when teams are tied in the third period than those same teams produced in the first. Bruce has done a study on the effect of the three-point game and I think that this data confirms his suggestion that teams are playing less aggressively in order to get to overtime.

I also thought it would be instructive to look at the different game states by breaking down the situations even further, both by home vs. road and by the amount of the deficit, i.e. one or two goals. These are the results when the teams are tied and when the home team is in the lead:


The change is significant both with a one-goal lead and with a two-goal lead. Now let's look at the results when the road team is leading:


As expected the road team also plays to the score. The difference in the results when the road team is up by two goals are particularly astounding. They outshot the home team in the first, presumbably to get a lead and were then run over territorially in the third as they protected the lead. The difference is far more pronounced than the same situation when the home team has the lead. This may be the result of the matchups. Of note, however, is that these totals represent only 13 games where the road team is up by two and only 12 games where the home team is up by two.

One final point. Certain teams compete in more of these games than others. The average number of games is 9.5. The highest totals are the Minnesota Wild with 15, the Tampa Bay Lightning with 14 and a few others at 13. The lowest totals are for the Detroit Red Wings with 3, the Columbus Blue Jackets with 6 and the Chicago Blackhawks with 7. The Oilers competed in 8.

2 comments:

Kent W. said...

Thanks for this Scott. Interesting stuff.

Scott said...

Thanks Kent. I think it just pinpoints some of what's already being observed. We already observed that tied games become more conservative and that teams with a lead would play rigorous defense but hopefully this provides a reasonable guess at how much.