Balls In Play Study

[Crash Course]

What's he saying here?

http://bradbury.sewanee.edu/Balltypes.pdf

[Makofan]

I don't know - my eyes glazed over

[Crash Course]

Me too.
Then the circus music started in my head.

[mainsr]

See, guys, this is why you have an applied math major in the group...

There was a study a few years ago by a guy named Voros McCracken that hypothesized that the only things over which pitchers have control are strikeouts, walks, and home runs. Everything else, the study hypothesized, is luck. (I shouldn't just say "hypothesized," there were data to support the conclusion as well.) What that suggests is that when a ball's put in play, whether it's a hit or an out is mostly a matter of luck, not pitcher skill. That means that based on a pitcher's IP, HR, and K, and his team's defensive effciency rating (the % of balls in play that are hits), you can estimate the number of hits he'll give up. If the actual hits allowed varies a lot from the estimate, it's a fluke, and the guy's likely to reverse that the following season.

In other words, take these teammates:
A 0.7 HR/9 IP, 3.1 BB/9 IP, 9.9 K/9 IP, 6.1 H/9 IP
B 1.0 HR/9 IP, 3.1 BB/9 IP, 2.7 K/9 IP, 11.1 H/9 IP
Now, A is obviously the better pitcher. But if you look at hits excluding home runs divided by IP excluding strkeouts, A has 8.56 and B has 11.28. McCracken's conclusion would be that Jason Schmidt (A) is obviously a better pitcher this year than Kirk Rueter (B), but a good chunk of the 5 hits per 9 innings differential is luck, not skill.

Don't jump on me, this is McCracken's study, not mine. See the Tommy John entry in the New Bill James Historical Baseball Abstract for more on the topic.

What this guy tries to do is figure out whether there is more to the story that this. He adds three other variables: percentage of batted balls that are line drives, groundball/flyball ratio, and percentage of fly balls that are hit to the infield. What he finds is that, in attempting to predict ERA:
1. % of line drives is somewhat significant, G/F less so
2. % of flies that are IF is not significant
3. None are as significant as the three items identified by McCracken

The new study has a serious limitation, as I see it: He uses data from only the first half of the current season. That's a very small sample.

This article's conclusion: If you want to predict a pitcher's ERA, look at his HR, BB, and K ratios and his team's DER. Hits allowed, other than home runs, is largely a function of luck, although there may be some skill--mostly in avoiding line drives and ground balls--involved as well.

[Crash Course]

Well done Rob. Thanks.
Goes to show, you can have all the brains in the world and figure out something very well - but, having the skills to communicate it in a manner that is easily understood is a whole 'nother animal.
That latter skill is something you've been blessed with!

[KCBOOMER]

There has been considerable work done in this area that elaborates on McCracken's work. The first thing is that McCracken's data set was some what flawed. He only used a couple years worth of data. When all the data available (100 years) is used we do find that some pitchers seem to have the ability to control BABiP (Batting Average Balls in Play). These pitchers are few and far between but they do exist.

The current thinking seems to be that even accounting for these outlier pitchers that BABiP is a function of luck and/or defense. The wizards studying this area seem to charge a given pitchers deviation from the league average BABiP to luck, but a given team's deviation as good defense.

BTW in mainsr's example above Schmidt's BABiP is 5 / 22 = .227 while Reuter's is 10 / 35 = .286. I'm doing this from memory but the formula is (H - HR's) / (AB's - HR's - K's). Neither a home run nor a strike out is part of the equation. And like regular BA neither walks, HBP's, nor sacrifices count in the equation.

Thus you can see that the difference in Schmidt's hits allowed in a BABiP world is not 5 less than Reuter, but more like 2. Reuter is around the league average, but Schmidt is still well below. Thus Schmidt has to be "lucky" because they both pitch in front of the same defense.

[mainsr]

Thanks for the clarification/correction, Boomer. I used IP instead of ABs, and I forgot to subtract HRs.

The fact that there is anything at all to McCracken's analysis is pretty provocative, IMHO. One of the most interesting aspects to me is that it suggests that if a pitcher's hits allowed are out of whack one way or another one year, they're likely to revert to the mean the next...a great indicator of who is likely to improve or decline from year to year. Great stuff for a GM, whether in the MLB or Rotisserie.

[mainsr]

Originally Posted by Crash Course

Well done Rob. Thanks.
Goes to show, you can have all the brains in the world and figure out something very well - but, having the skills to communicate it in a manner that is easily understood is a whole 'nother animal.
That latter skill is something you've been blessed with!

Thanks, but let's be fair to the author...The piece you lifted was written in academia-ese; the guy's a college prof. He also has a blog that contains a much better synopsis, albeit strewn with undefined abbreviations:

In the paper, I try to analyze the impact of batted ball types on pitchers' BABIP and ERA. DIPS/FIP theory says that pitchers have little impact over balls in play; however, there have been a few studies that find some role for pitchers to affect certain batted ball types. In particular, line drives,infield flies, and the ground ball/fly ball ratio seem to be things pitchers can control. Since THT keeps tabs on all of these stats, I decided to quantify the impact of these batted ball types. Here is a brief summary of the results.

Both LD% and the G/F ratio both affect the likelihood that a ball put into play will be hit; furthermore, these factors impact pitcher ERAs. However, the impact of these factors is much smaller than the three main components of DIPS. We know that pitchers can control walks, strikeouts, and home runs from year to year, and that these factors are significant components of ERA. Without including hit ball types, these three factors explained over half of the variance of ERAs of pitchers for the first half of the 2004 season. Adding the hit ball types does very little to further explain the variance of ERAs.

BABIP - As Boomer noted, batting average on balls in play
ERA - If you don't know this, you need to find that preseason football blog you were looking for.
DIPS - Defense-Independent Pitching Statistics
FIP - Fielding-Independent Pitching (same as DIPS)
THT - The Hardball Times
LD% - % of batted balls that are line drives
G/F - groundball to flyball ratio