Sunday, May 13, 2007

Run Differential, the Neyer Way

Here's the garbage. Here's the garbage man:

"NFaleris (5/11/2007 at 4:28 PM)

I think the bigger issue is that run differential is better used as an analytical tool (e.g. for isolating "luck") than as a predictive tool. Not only will blowouts (like the A's and Rangers yesterday) affect the predictive capacity, it will have an ABSURD effect this early in the season. While I agree with most of what you write, Rob, it's a little silly to think the run differential after the 24th game of the season would tell us the Cubs would be a potential team to beat -- which is essentially what you say above."

(temp link, look for "Run differential Matters" Rob Neyer May 11th, 2007)

To add to NFaleris's great comment, the Blue Jays had a +23 on that date as well, and were 13-12. Now they're 15-21 with a minus 20 differential. Run differential is analytical to see how a team is performing, not necessarily predictive without a huge sample (but then again what isn't predictive after a huge sample?). That's what I was getting at back in "The Illusions of April." I should probably email that blog to Rob again, I don't think he read it too closely.

Personally if I try to predict a hitter's future performance, I look at OBP - AVG, and SLG. If the guy has OBP - AVG at 0.07 or higher I predict good things, same for SLG over .500. For pitchers I look at K/BB only. About 3/1 ratio and higher I predict good things. This baseline is pretty much based empirically on Webb's past performance, since he's the most extreme contact pitcher who is actually good, and that's about what his K/BB is generally. I think the same stats are good predictors for teams, but also record in 1 run games is relevant since a team with a very good 1 run game record will most likely fall off their pace and vice versa. In this way you can avoid the skewed data of run differential where a couple blowups from fringe pitchers can impact it.

Caveat: Some weird things happen when players get old. Old hitters often increase their walk rate as their bats slow and they foul more pitches off instead of putting them in play. This brings up their OBP - AVG. Since higher OBP helps their team it's still not a bad predictor of "good things." But on the flip side, as great pitchers age, they can maintain a high K/BB as they lose effectiveness. I hypothesize that this is because "stuff" degrades before control, so a guy could continue throwing strikes as his stuff gets worse, leading to more hits and runs scored against. That's not a "good thing", but anecdotally it only seems to happen to the great ones who refuse to paint the corners even as their stuff diminishes, since they've never pitched that way before. See Schilling (2005) who seems to have adapted or Johnson (2006-) to see what I'm hypothesizing about.



Blogger Jumpkin said...

Please explain Chien-Ming Wang.

His K9 last year was 2.9!
K/BB 1.46!


12:55 PM  
Blogger sabesin2001 said...

His stats other than era certainly don't predict greatness. What it really comes down to is that either Wang will regress to a very average pitcher, or we are witnessing the best groundball pitcher modern baseball has seen.

But if he was the greatest groundball pitcher it would show in his groundball/flyball ratio. For comparison, Brandon Webb has an absolutely sick gb/fb ratio of 3.85 for his career. The good Derek Lowe has a 3.34. The average Jake Westbrook has a 2.72. And the bad Carlos Silva has a 1.56. All are labeled as groundball pitchers who don't walk many and don't strike out many. Now for Wang, his ratio is 2.99. Better than Westbrook but not as good as Lowe.

As for homers/9 innings, Lowe is 0.7, Westrook is 0.8, and Wang is only 0.6. That is even better than Webb somehow, who is at 0.67.

The conclusion to draw from this is that Wang gets more fly balls than the excellent ground-ballers, but at the same time allows less homers.

Since I don't believe that there is such a skill as inducing medium-deep fly ball outs, the most reasonable conclusion to make is that Wang will eventually converge on an era in the mid 4s, a fairly average pitcher who doesn't walk many.

11:21 PM  
Blogger Jumpkin said...

It's actually interesting to compare his 2006 to Lowe's. They put up essentially the same numbers except Lowe struck out 47 more batters altogether (1.7 more per game.) Wang walked about the same number of batters, but slightly more hits. They finished with the exact same ERA on the year too, but Wang's PERA was higher.

I guess my point is that it's really obnoxious that I have to hear him described as New York's ace all the time just because of the big 19 in his stupid W column last year (which naturally made him 2nd in the Cy Win Competition.)

9:52 AM  
Blogger sabesin2001 said...

2 things i am thankful for from last year were Wang not stealing the Cy Young and Cano not winning the batting title. something to be thankful for this year: Wang not throwing a perfect game whewww.

what was Wang's PERA last year? like low 4s?

10:05 AM  
Blogger Jumpkin said...

Yeah I actually saw the Wang near-perfect game. I tried hugging Broussard through the TV but it didn't work.

2006 PERA = 3.87.
2005 (partial season) PERA = 4.39

11:54 AM  
Blogger sabesin2001 said...

yeah that was huge, Broussard was so fooled on that pitch too.

6:48 PM  

Post a Comment

Subscribe to Post Comments [Atom]

<< Home