Thursday, November 09, 2006

Wins v. ERA

With the World Series over and free agency still a few days away, there's not much going on in the baseball world. I figured I'd take this time to quickly go over the uselessness of wins for pitchers.

Proponents of wins insist that some pitchers "know how to win" and their records supersede their ERAs. Opponents, on the other hand, say that pitchers can't control their run support. Now it occurs to me that this is very easy to check.

Consider: two pitchers, one with 13 wins and a 3.50 ERA and the other with 18 wins and a 4.50 ERA. Who would you rather have next year?

Let's find out. I decided to take the top 20 pitchers in MLB in 2005 in wins and ERA, and see what they did in 2006. Due to ties, we end up with the top 23 in wins and top 24 in ERA. Here are the average results for each group:

Wins leaders: 17-9, 217 IP, 3.57 ERA, .645 Win%
ERA leaders: 15-9, 212 IP, 3.10 ERA, .629 Win%

Both groups were obviously very good, in large part because ten pitchers appear on both lists (Willis, Colon, Carpenter, Oswalt, Garland, Pettitte, Santana, Buehrle, Beckett, Martinez). So let's remove them:

Wins leaders: 16-10, 211 IP, 4.04 ERA, .617 Win%
ERA leaders: 12-9, 203 IP, 3.18 ERA, .581 Win%

That's more like it. Nearly a full run difference in ERA one way and a four-win difference in their records the other way. Present those two pitching lines to baseball fans and you'll find quite a divide in their responses.

Fast forward one year. This is what our two groups did in 2006:

Wins leaders: 11-11, 177 IP, 4.59 ERA, .509 Win%
ERA leaders: 12-10, 182 IP, 4.18 ERA, .554 Win%

How do you like that? Not only did the ERA group continue to post a better ERA the following year, it also had more wins, fewer losses and more innings. And that's despite Roger Clemens sitting out the first half of the season.

Bottom line: this year's ERA is actually a better predictor of next year's W-L record than this year's W-L record is. Which means judging a pitcher by his win total is silly.


sam said...

I am a little bit worried that the ERA leaders suffered more significant drop-offs the next year compared to the win total guys. It seems they had the more flukish season. But I guess that is a topic for another day.

Jeteupthemiddle said...

That's what I was wondering about, sam. The ERA of the second group went up an entire run between 2005 and 2006, where as the ERA of the first group went up half a run.

The ERA group was still better, but not nearly as good.

Perhaps I will look into other years.

I have no doubt that ERA is a better indicator of a pitcher's value than wins, but it is still interesting to see.

Anthony said...

A lot of that is simply regression to the mean. All players (and teams) tend to move back to average. And the pitchers with the best ERAs have farther to fall back.

Luck is a factor in that obviously. As Sam said, there will be flukes involved in this sort of thing.

The biggest ERA dropoffs:
Carlos Silva (3.44 to 5.94)
Kevin Millwood: 2.86 to 4.52
Jarrod Washburn: 3.20 to 4.67

All those seem like guys who were lucky in 2005 and unlucky in 2006. After that comes Tim Hudson, Joe Blanton and Jake Peavy. I'd say Hudson and Peavy were normal in 2005 and very unlucky in 2006.

Of the top 24 in ERA in 2005, only one improved in 2006 (Johan Santana: 2.87 to 2.77).

sam said...

Actually, that was what I was getting at, and said that was a topic for another day. It ties more to the Hardball Times style peripherals argument.

Point is, both are relatively bad predictors of performance, but ERA is lesser of the two evils.