Bad Statistics

There is an article in the January 31st issue of Sports Illustrated on the emergence of Ryan Kesler as a premier player in the NHL (read it here). If that is really true or not could likely be the subject of another entry, although his 18% shooting percentage is sure making him look more offensive than he really is. It the article contained one line that I found particularly irksome. It is a statistic that is thrown around quite frequently and one that has never seemed all that meaningful to me.

“As Kesler goes, so go the Canucks. Through Sunday, Vancouver (29-10-9, tops in the West), which has lost only three games in regulation since Dec. 5, is 24-1-5 when Kesler gets a point.”

You see this kind of number thrown out all the time. A certain team’s record is much better when a certain player scores as opposed to when they don't. The point seems to be that player x is particularly key to their team’s success. Looking at the numbers, it turns out saying this about basically any player will make you correct.

Since the article was about Kesler, I decided to look at the Vancouver Canucks. I logged who scored in all of their games through their February 7th win over the Ottawa Senators. I looked at the point percentage for each player in the games they scored, versus the team's overall point percentage. As expected, the team has a better record nearly regardless of who scores. Of the 27 players that have scored for the Canucks, the team’s record was better for all but 3 players. The outsiders: Peter Schaffer (16gp), Sergei Shirokov (2gp), and Keith Ballard (53gp, but only 4 points).

Here are the full results:

In hindsight this is beyond obvious. First, any game where Vancouver is shut out gets automatically eliminated Second, in it’s easy to understand goals are more plentiful in games won than lost. Teams don’t often lose when they score 3 or more goals. Teams almost always lose when they score less than that.

It also turns out that Kesler doesn't even have the most impressive results when looking at these numbers. That honour goes to Mason Raymond, with the Canucks going 17-1-0 in games where he scores. Instead, it suggests that the stat is completely meaningless.

I also think this does a disservice to a player like Daniel Sedin. He has scored at least one point in 81.5% of his team’s games. This kind of consistent performance is much more beneficial to the team over the course of the entire season.

I would conclude that that trying to directly correlate any individual stat to any team win is probably a bad idea. The next time you see a stat like this quoted, it should definitely be taken with a grain of salt.