Thursday, May 8, 2008

Which teams overachieved this season?

This is a bit long-winded so please bear with me.

You know how Bill James figured out that a teams won/loss record was directly connected to its total runs for/runs allowed for the season? Well there's no reason that the same thing shouldn't apply in hockey. In fact, if you graph the ratio of each team's goals for to its goals allowed against its points total for the past 3 seasons you get this fun little formula: 76.73X+13.27, where X=(total gf/total ga). Punch in your team's value for X and you should get a reasonable estimate of its points that season.

The principle behind this is pretty simple (it works the same for the ratio, but i'll use the difference between gf and ga in the example coming up). If team A outscores its opponents by 0.5 goals a game, it should do better than team B, who gets outscored by 0.5 goals a game. Luck may fuck with this of course. For example, team A could win one a game by 5, then lose the next three by 1 each, while team B could lose a game by 5, then win the next three by 1 each. In this situation team A stilll outscores its oponents by an average of 0.5 goals a game (5-1-1-1)/4=0.5, while team B is still gets outscored by 0.5 goals a game, but nonetheless team A goes 1-3-0 while team B goes 3-1-0.

If you assume that such variations are due to luck (i.e. that teams can't purposely apportion the goals they're going to score across different games as it suits them), then over the long run these breaks should even out, and a team's record should end up quite close to the record predicted by the formula. If a team's record is a lot better than its goals for/allowed would indicate, this is probably due to luck rather than some intrinsic ability on the part of the team to apportion its goals. Thus, the team is overachieving and its results should get worse as its luck evens out. Indeed, the four biggest "overachievers" last season based on this formula were Vancouver, New Jersey, Dallas, and Atlanta (all had many more points than the formula would indicate) and all 4 took a step back points-wise this year. (Of course it doesn't always work. The formula said Pittsburgh overachieved last year too.)

So who overachieved this year and is due for a step back next season? Without further ado here's what each team's record would have been this year based on the formula above, compared with its actual record.


No comments: