Bill James’s “Pythagorean Expectation” for estimating baseball teams’ win percentages is one of my favorite sabermetrics. The formula is simple and powerful:
It is also remarkably accurate. It would suggest a 42% win % for the Mariners so far this season, and they’re actually 11-14 (44%). Not bad. However, for a team like the Mariners that is built on pitching and defense, it doesn’t always hold water. For instance, last year the model suggested 75 wins, but the team actually won 85.
While watching Mariners games this season, I wondered if a team that is so hard up for runs is more affected by errors and if that statistic could be a good predictor of winning. Looking at the 25 games so far this season, it turns out that error differential (opponents errors – Mariners errors), is also a pretty good predictor of win %:
The model (R^2 = .75) suggests that for each additional error that the opposing team makes over and above the Mariners, the Mariners’ win % increases by 20% (absolute). Results have shown the model to hold up fairly well:
- Mariners have not won a game while making 2 errors more than opponents; similarly, they have not lost a game when their opponents have made 2 errors more than them.
- 1 additional error by the Mariners = 20% lower win % than 50% (29%)
- When errors are equal, win % is near 50% (although the model suggests closer to 40%)
- The one place it doesn’t hold is when the Mariners have the benefit of 1 error, they haven’t won as much as the model suggests – 33% versus 60%
Knowing this, I wonder if the move to send Tuiasosopo back to the minors was driven by his fielding. He leads the team in errors (3) and has only started 5 games.
I’ll keep an eye on errors this season and see if the model holds up.
Pythagorean Expectation graphic source: Wikipedia