With the off-season in a holding pattern, let’s spend a little more time talking win values, shall we?
One of the questions that came up in a thread last week was how well team win values match up with actual wins. So, I wanted to spend some time on that issue. The first thing to remember is that our win values are context neutral, so we are not attempting to account for the distribution of runs in regards to the value of the situation. In the win value system, a run in the 3rd inning of a 10-0 beatdown is the same as a 9th inning walkoff that decides the winner. Because we’re presenting context neutral wins, teams that have a run scoring distribution skewed towards higher leverage situations will win more games than win values would suggest.
The Angels are the best example of this. Last year, they accumulated 17.2 wins from their position players and 22.0 wins from their pitchers for a total of +39.2 marginal wins. With our replacement level set at .28852 (or just round to .289) for ’08, that makes the Angels an 85.94 win team according to our system. Basically, we think they earned 86 context neutral wins last year. They actually won 100 games – 14 more than their win values would have suggested. Why? They were the Kings of Clutch, performing drastically better in situations that mattered than they did when the game wasn’t on the line.
If you’re familiar with the concept of Pythag Win%, you’ve seen this phenomenon before. Because there’s little to no evidence that the distribution of run scoring is a repeatable skill, we’re not including it in our win values, and will match up with pythag wins better than it will with actual wins. This is born out by the correlations of each.
Win Values to Pythag: .90
Win Values to Wins: .85
Win Values correlates very well with both pythag wins and actual wins, but better with pythag because of the run distribution issue that we talked about above. Clearly, though, our system of rating players is adding up very well at the team level, even without considering the context of when runs are scored.
This evening, we’ll take a look at the differences between win value projections and pythag expectations and explore why two projected win totals both based on runs scored and allowed would diverge.