Monday, July 30, 2007

It's All About the Team

Over the last few weeks, I haven't posted much, as I have been working on a stats project, developing a new metric to evaluate historical NHL goalies. In recent years, there has been a big improvement in our evaluation of goaltenders, first as a result of a movement away from team-dependent stats like wins and GAA and towards save percentage as the key metric, and then further refining that by adjusting for era or workload. This is definitely a step forward, but even that ignores probably the most important single factor in a goaltender's success: The team in front of them.

Perhaps the biggest breakthrough has been shot quality measurement (read up on that at Hockey Analytics), which has allowed us to remove the impact of the defence in front of a goaltender, subject of course to the limitations of the NHL's RTSS data. However, these data are only available for the past 5 years or so. How then can we attempt to correct for team factors for goalies throughout the history of the NHL?

We can't, really, at least not with very great precision. However, there are ways to estimate shot quality. I looked at correlations with various variables and team shot quality, and found that the variables that were most closely correlated were overall goals against, and team save percentage. Therefore, to estimate shot quality, we can use the statistics of a goalie's teammates, and by using that as a benchmark, evaluate whether the goalie's performance is exceptional or team-driven.

There are several issues with this, primarily sample size. The elite starting goalies play 70-75 games a year, meaning sometimes there are only a few games played with someone else in the net. Also, I don't have save percentage information from before 1982, so I used goals against average. GAA is very team dependent, but since we are comparing teammates here I consider it to be reasonable to use. I also am not sure how much I can trust shot totals over the years, and according to some theories certain goaltenders are responsible for reducing the shots against themselves through puckhandling, rebound control, etc., so using GAA removes those potential issues. Backups also tend to be weaker goalies, but they also tend to player weaker opponents as well. To adjust for this, I have a subjective method for adjusting backup performance, based on their ability and career record.

In short, then, if a goalie gives up fewer goals than the other goalies on his team do, he is better than them, and is contributing to the team's defensive success. If a goalie has similar stats to the other goalies on his team, then it is likely the team that can be credited with the defensive success. In general, the greater the gap between the goalie and his teammates, the better the goalie.

My main area of interest is to determine who were the best performing goalies of the 1980s - was it Fuhr, Smith, Roy, or Vernon, i.e. the goalies with the rings, or were there other unsung heroes who never received their due? This project is almost completed, so I will be posting some of my findings, especially the cases of goalies who appear to be substantially under- or overrated, and hopefully expand this blog's dialogue to include many other goalies beyond Martin Brodeur.