[Jason Moyer]

I think understanding the system is easier if you do two things: 1. realize that it's not perfect, but it's the best we have available and 2. look at it in outline form

Basically here is a good summary of how you figure win shares. I'm not going into the different ways that he weighs things, but you'll get my point:

1. Start with team wins
2. Multiply times 3 - this is a team's win shares.
3. Divide this between offense and defense
4. Divide the offensive win shares between players based on their runs created per out.
5. Divide the defensive win shares between pitching and defense.
6. Divide the pitching win shares between pitchers based on how many runs they saved compared to norm, with a slight weight given to win/loss record and saves.
7. Divide the fielding win shares between fielding positions.
8. Divide the fielding position win shares between individual fielders based on weighted stats.

The offensive win share concept I find pretty straightforward. Teams win by scoring as many runs as possible in 27 outs. The runs you create for your team vs. the number of outs you create is a great indicator of your relative worth to the team, with your bat.

The pitching win shares are a bit more complicated. Essentially, a pitcher is given the most credit for striking out people, not allowing home runs, and not allowing walks with some weight given to ERA and W/L/SV. The win shares system assumes that while a pitcher does have an effect on the number of hits he gives up, the defense should also be given credit for those hits. Greg Maddux will allow a lot more hits with a weak defense than he will with a good defense, it's common sense. The things a pitcher has the most control over are K's, BB's, and HR's.

Fielding is perhaps the most complicated, and where most people are likely to disagree with James. However, I think this portion is where he actually contributes his most groundbreaking work ever.

Just to give some examples, catchers are generally credited based on how many runners they throw out, how many assists they record, and how many independent putouts they record (i.e. putouts that aren't strikeouts, since catchers are credited for those as well). There is a ton of insight into both the individual performances of catchers using this method (Johnny Bench's statistics are incredible if you remove strikeouts from his fielding stats) and also how the position itself has evolved over time. Since 1900 the number of independent putouts by catchers in proportion to other putouts has decreased steadily. Why? Are there fewer bang-bang plays at the plate? Fewer popups fielded by the catcher? Who knows. But with this information we can start to really analyze how the numbers depict the game.

Other interesting defensive information that he focuses on... First baseman's independent putouts, i.e. putouts that aren't the results of assists by other infielders. First baseman's independent assists, i.e. assists that aren't the result of quick flips to the pitcher.

Further, the beauty of this is, there is finally concrete proof that a fielder's raw total chances has nothing to do with how good of a fielder he is. By comparing a fielder's statistics to the total balls in play and the situations in which those balls in play are occuring (i.e. are there a lot of men on first? Maybe that's why there are so many double plays!) you start to unravel the puzzle of fielding stats, which are normally meaningless.

My favorite part of calculating Win Shares is going through the tedious bits where you determine expected stats. For instance, given a certain number of balls in play, and a certain ground ball ratio, and a certain number of men on first base (all of these estimated using normal fielding stats) you can estimate how many double plays a team should have made that year, assuming their double play ability was average.

Interesting interesting stuff. If you're having a hard time grasping it, sit down with a notebook someday and make an outline of the formula. It's a lot more straightforward than it appears.

Jason

<small>[ 06-04-2002, 07:22 PM: Message edited by: Jason Moyer ]</small>