by PuckStopsHere on 10/20/09 at 01:33 PM ET
The Edmonton Oilers blogs include some of the better sabermetrics blogs out there. There is mudcrutch hockey, irreverent Oiler fans, copper n blue and battle of Alberta. Each have a strong interest in sabermetrics and have written some interesting posts on the subjects. With the high quality of the Oiler blogosphere it is hard for the mainstream media to keep up. However, Dave Staples at Cult of Hockey does a good job. He writes many interesting pieces about the Edmonton Oilers and hockey in general. These are not just the leftover articles that do not get published in the newspaper (as some weaker media blogs tend to do). He regularly publishes some very good information on the internet.
However, one of his recurring themes (that seems to be in protest to the independent Oiler bloggers with strong sabermetric backgrounds) is that hockey statistics are not very useful. This is a case he frequently overstates in his posts. Last year he wrote Why Plus/Minus is a rotten, useless, misleading and irrelevant stat for NHL players.
This title is overly strong, but he basically argues that +/- is problematic because it depends upon context. For example, a player on a good team will get a better +/- than one on a bad team regardless of the quality of the players in question. He also argues that sometimes players get pluses when they didn’t do anything significant to contribute to a goal or minuses when another teammate screwed up allowing a goal and they didn’t do anything wrong. While all this is true, it hardly makes +/- useless. It merely makes it important to keep context and the random error in +/- numbers in account when trying to argue using +/- to back up your argument. I wrote a responce to his rant here .
Nearly a year later, Dave Staples is back with another rant entitled what good are hockey statistics if they don’t even tell us who scores and causes goals? This title clearly overstates his argument again. Hockey statistics clearly do tell us who scores goals. The number of goals a player scores is the oldest statistic kept. The problem he is trying to elude to is that one cannot clearly and unambigously show how many goals a given player creates for his team over a season. There are crude estimates of goals created but they do not adequately address the issue. The problem is that hockey statistics do not give all the information necessary to understand a hockey game - and it may be impossible to gather enough information to ever fully understand everything statistically. That does not mean that considerable understanding cannot be found statistically. That does not mean that significant advances in understanding are not on the horizon. It only means that sabermetrics is only part of the picture toward understanding hockey. It is a useful tool that one cannot throw away because it is not a 100% fully developed, fully proven tool.
Dave Staples goes on to an interview with somebody named Corey Pronman who is attempting to do something about this statistical problem, but is probably headed in the wrong direction. Mr. Pronman has introduced a subjective element into his statistics. He watches replays of gals and subjectively choses the players who he thinks deserve credit for the goal or blame for the goal and gives them a subjective plus/minus rating he calls individual plus/minus or real plus/minus. The nature of this statistic is that it will inevitably show which players are best at doing whatever Mr. Pronman thinks is important in scoring goals and which players regularly do what Mr. Pronman thinks cause their team to allow goals. That subjective element cannot be removed. Everything is seen through Mr. Pronman’s filter. That in and of itself is not a bad thing, but we cannot learn anything about which players are important that isn’t already built into Mr. Pronman’s subjective evaluation of goals. The subjective element of statistics needs to be removed and reduced as much as possible to truly learn about the underlying elements of the game.
The main problem with hockey statistics is that not enough information is gathered. In a typical game, six or seven goals are scored. We record who scores and assists on those goals and who was on the ice for +/-. That is it. The numbers you see for a player in a typical box score are based on tallying an average of six or seven events in a game. Roughly one event for every ten minutes of playing time is tallied. That leaves significant information out of the analysis. The effort to fix this is to record more information. That is the effort behind the Corsi number . In Corsi, each shot directed at the goal is included (that includes shots on goal, missed shots and blocked shots). This gathers far more entries into the dataset for evaluation of players. Perhaps all of those elements are not equal. A goal scored is clearly more important than a missed shot. Somebody in the future may find a better way to tally all of these events (and find other meaningful events that can be tallied with a minimum of subjectivity) to give a better dataset in which to analyze players contributions. From this dataset, one can work out better and better models of the contributions of players and be better able to quantify the value of a given player. That is the goal of hockey sabermetrics.
I think the idea of subjectively going over statistics to award those players one thinks did the right thing and penalize those who didn’t when goals are scored is not the right approach. The approach is to build as large a dataset of significant events and find the appropriate objective analysis of those events. It is not to add in individual bias in scoring by making it more subjective.
To answer Dave Staples question: What good are hockey stats if they don’t even tell us who scores and causes goals?, hockey stats do a good (but imperfect) job of telling us who scores and causes goals and that clearly shows that they have a strong value. With future understanding, hockey stats can be made even more useful than they are today.
Be the first to comment.
Add a Comment
Please limit embedded image or media size to 575 pixels wide.
Most Recent Blog Posts
About The Puck Stops Here
Who am I? A diehard hockey fan.
Why am I blogging? I want to.
Why are you reading it? ???