Wednesday, August 27, 2008

Spoiled, Obsessed and Fuming

Spoiled, obsessed and fuming is no way to go through life. Or at least not through a baseball season.

Talking about the Eddie Guardado deal to several friends, I remarked how it crystalized just how pathetic the Twins bullpen situation really has been. The Twins picked up Guardado, who is really nothing more than an average reliever at this point in his career, and he's instantly the second best guy on the staff. And maybe the only other guy in which the fans might have some confidence.

If that doesn't clarify exactly how bad the bullpen has been, nothing does, right? We had Nathan and five to seven guys who were - at best - below average. So why the hell did it take so long to address this? What little delusional hole does Twins management find themselves in.

It was a great point - had it not been completely wrong.

On two Sundays this month, I have literally fumed for hours for at the Twins inability to upgrade their bullpen. Some would say that's a sign that I don't have a life, and I readily concede that. But, to be fair, in both cases I was returning from a road trip and precious little else to obsess about.

But the Twins bullpen pitchers aren't bad across the board. has a slick little statistic it tracks for relievers called Average Run Prevented(ARP). It supposedly measures how many runs a reliever prevents versus your average pitcher. (How it does this is kind of neat, and for you other geeks, I'll cover it at the end of the story). So if you have a positive factor, you're an above average reliever. Negative? You're below average. Going into yesterday's tilt, here's how the members of the Twins bullpen ranked:

Joe Nathan +20.8
Craig Breslow +9.2
Bobby Korecky +5.7
Jesse Crain +5.4
Matt Guerrier +3.0
Dennis Reyes +2.7
Pat Neshek -0.9
Brian Bass -4.3
Boof Bonser -4.5
Juan Rincon -13.0

The existing members of the bullpen, with the exception of Boof, are all above-average relievers this year. And if you're wondering, Guardado's ARP for the year is 8.6, which means he's not quite the second most effective pitcher on the staff, or even the best left-hander. Breslow beats him out.

This doesn't mean that Bill Smith is off the hook. I can still list an even dozen realistic things the Twins could have done to help out their bullpen this year, and none of them were done. In fact, between 5/29 (when the Twins claimed Breslow) and this week - or for almost three months - the only enhancement the Twins made to the bullpen was to subtract members from it.

But it does mean that I'm probably overreacting, which is apt to happen during a pennant race. Perhaps the bullpens in previous years might very well have spoiled us. Or perhaps the bullpen is better this year than I've generally given them credit.

Whatever. I'm still not sure I totally believe it. But it will help me make it through the season.


Ok, for the fellow geeks, here's how ARP works...

You start with a list of the average number of runs that score in a given situation, based on historical MLB data. For instance, a reliever comes into a game with runners on the corners and one out. The average number of runs that score in that situation is approximately 1.14 runs.
  • If he escapes that inning without letting any runs score, he gets 1.14 ARP, because that's the average number of runs he prevented.
  • If he escapes that inning letting just one run score, he gets 0.14 ARP, for the same reason.
  • If two runs score that inning, he gets -0.86 ARP, because he let in more runs than average.
  • Finally, if he gets one out (no runs score) and is then replaced by another pitcher, the average number of runs that score in the new situation (runners on corners, two outs) is approximately .46 runs. So he gets credit for the change of 0.68 ARP (because 1.14-.46=.68).

It's a valuable stat because it gets rid of all the inherited baserunner crap that relievers usually need to deal with. But it's important to note that it's value is that it tells you what has happened, and isn't terribly valuable as a predictive stat.