Friday, October 03, 2008


Junk stats

A brief side note: my Stats Primer project thingy is still ongoing, but it's going to take a while. I've had half a post on junk stats and skill vs. luck written for a while, but getting these things "right" has proven more challenging than I hoped, and also a little bit too much like work. It'll happen eventually though.

I bring this up because the first paragraph of that post is on what I think is probably the worst kind of stat, and the kind that turns people off of paying attention to new ones altogether: the stat that is advertised as measuring Thing A, but is massively affected by Things B, C, D, and E.

The common stat that's the best example of this is ranking goalies by their raw Win totals. For a goalie to get a Win, his team needs to create shots and scoring chances, which he can't control. He also can't control the performance of the guy in the opposite net. He can't control the number of shots or the quality of scoring chances his own team allows on him (with the marginal exception of rebounds). In a meta sense, he can control how many starts he gets (generally play better --> get more), but it's still someone else's decision, which will also be affected by the coach's philosophy, the schedule, and the quality and health of his backup. Basically, using # Of Wins to assess goaltenders leaves almost everything to be desired.

Fortunately, we have The Hockey News around to make everything much worse:
...we at The Hockey News devised the Goalie Confidence Index or GCI: an objective formula concocted to measure how effectively a team can rely on its netminders.

Tracking both starters and backups, we ranked the NHL’s 60 goalies based on save percentage, goals-against average, save percentage on the penalty-kill, percentage of games in which the goalie was pulled, percentage of games giving up five goals or more, winning percentage and shutout percentage.

By adding together a goalie’s rank in each category, we came up with a score.

Obviously I don't understand who THN's target audience is. I don't understand why they used a 2-page spread available for a feature on goalie statistics and, instead of doing something that might actually inform their readers, invented a system out of thin air that means nothing and is guaranteed to be forgotten by the time the page is flipped.

I don't understand why 5 goals or more was chosen as the cutoff number for a bad game. I don't understand why pulling the goalie is used at all, when the #1 reason is not the amount or even quality of goals he has allowed, but the difference on the scoreboard. I do understand why the "rank, then sum" method is used (it's easiest!), even though it has umpteen inherent problems. I don't understand why these 7 categories of varying importance were all given equal weight.

I dunno. It's their magazine and they can do what they want. I know they have some good feature reporting on occasion, and I don't begrudge them their lists of the Ten Snappiest Dressed Coaches in NHL History, and the like. But the sizeable corner of the I bump around in consists mostly of people who want to understand more about hockey because it contributes to their enjoyment of the game, and no one ever says, "I discovered something interesting reading The Hockey News." It seems like the one niche that a print publication with a good brand has a chance of dominating, and they show no interest in filling it. I don't understand.


That was awesome Matt, great read.

I've got to admit, I would be kind of interested to hear the response of THN's editor to that post.

This comment has been removed by the author.

My guess is it would include the words "mom's basement" somewhere in there.

I remember THN's hockey pool guide a year ago had a line about some smart owner someday hiring a pool junkie as a GM because they know hockey stats better than anyway, and not just their favourite team, but the whole league.

I fully gave up on THN after that.

uh, that should be "better than anyone".

Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?