The Puck Stops Here

Since it is summer and there isn't much hockey news rolling in, its time to take a look at sabermetrics and hockey. The problem I want to look at is goalies. Goalies are harder to evaluate statistically then position players (at least offensive ones). In the past, I have shown Pnep's Hall of Fame monitor for goalies. It does an acceptable job of ranking the best careers for goalies although it is based largely upon awards won by the goalies and not based on a statistical analysis of their performance. It offers no way to predict who should win a goaltending award (such as the Vezina trophy) and is thus not fully satisfying.

The problem of goalie sabermetrics comes down to the poor choices of the standard goalie statistics that are kept.

One common statistic is goalie wins. Goalies do not win games. Teams win games. Sure goalies are part of a team, but does recording goalie wins make more sense than recording wins when a given defender leads his team in ice time? Last season, the top 13 goalies in wins all were on playoff teams. That is no coincidence. Good teams win games. A goalie on a weaker (non-playoff) team who may be good (such as Olaf Kolzig or Christobal Huet or Cam Ward) will not be a league leader in wins because his team doesn't win very often.

Another common goalie stat (arguably the most popular) is goals against average. The number of goals a goalie allows per 60 minutes of play may partially evaluate the goalie, but it also shows how good the defence is in front of a goalie. If a goalie faces lots of shots or high quality shots it is far harder to have a low GAA then if he doesn't. This statistic is also highly dependant upon the team in front of the goalie.

Shutouts are another common goalie stat. They record how many complete games a goalie played without allowing a goal. They say nothing about his result in the non-shutout games. Was he blown out or did he keep it close allowing 1 or 2 goals? Wouldn't the number of 1 goal or better games for a goalie be as meaningful as shutouts? Either of those stats would be highly team dependant since it is far easier to record a shutout when your team plays good defence in front of you.

The most meaningful, but still imperfect, statistic is saves percentage. It records the percentage of shots a goalie faces are saved. It is easier to have a good saves percentage if you face low quality shots instead of high quality ones, but at least the quantity of shots is normalized for.

I plan to write a few posts about attempts to provide more precise methods to evaluate goaltenders and evaluate their strengths and weaknesses during the remaining off season.