The Hardball Times has an article up today by Tango, part one of a series about his Leverage Index.  No one’s linked it at the real Primer yet, so I figured I’d put it here just so any interested parties know it’s out there, assuming they’re more likely to come to the Fora instead of just checking out the HBT every day.

I’ve been interested in Win Probability, but I’ve always thought a fault in it was that it’s equally weighted. The win probability of the Royals down by 1 in the second inning with two men on and one out should be different than the win probability of the Red Sox in the same situation.

Luckily, Tango at least began to address that in Recipe #3, the Great Hitter.

The problem with weighting teams or players is the smaller sample size. Also, where does it stop? Can you look at just the Royals, or a Royals lineup consisting of 9 specific batters in one specific order. If so, the bounds are increasing and the sample sizes decreasing. Regardless, I’d like to see some ways to weight certain players, teams, etc.

Others, like Tango and Keith Woolner agree with you.  But I don’t see why that’s a “problem.”  WP doesn’t have to be predictive.  It’s supposed to be a system that provides a benchmark to quantify each individual’s contribution to the win or loss.

It’s still a bit of a problem, even if it’s only providing that benchmark to quantify contribution.

For example, a Derek Lee home run does more for the Cubs than a Reggie Sanders home run does for the Royals, merely because the Royals’ pitching staff is more likely to give it back. To win on offense for the Royals, you’ll consistantly need six or seven runs, just to be safe. For the Cubs? Maybe only five or six to be safe. Also, Lee is in the NL, whereas Sanders is in DH-Land. Home runs are slightly more valuable when pitchers bat. It may be very marginal, but it’s a difference, and Sanders gets the same Win Prob. credit that Lee does for differently valued acts.

I think it all gets hung up in “contribution to the win or loss”—if it was just contribution in general, it might be okay to maintain the same benchmark from team to team, league to league. But if we’re measuring contribution to how the team did, the team needs to be somehow accounted for. If Derrek Lee were on the NL All Stars team, he’s almost as useless to them as a replacement player 1B, but if he’s on the AAA Iowa Cubs, he IS the show. He’s Babe Ruth in little league.

Is my thinking flawed here? This isn’t rhetorical. I have this sudden feeling that everything I know about Win Probability is wrong. :)