Daniel Cabrera... and BABIP

[furryburres]

Out of curiosity, where the world did you come up with the part in bold? I'm not saying it isn't true, only that I don't know that I've ever seen that written.

You also have to be careful with BABIP in that different sources use different forumlas. This year I have made it a point to use Baseball-Reference as my data source for just about every stat that I track. This hopefully keeps me more consistent. Their version of BABIP is:

(H-HR)/ (AB - SO - HR + SF)

I realize you're using Median and not overall average, but the overall AL BABIP this year is .291 (according to both B-R and THT). The overall Strand Rate or LOB% is 71% (according to THT).

You also really need to be careful with statements such as "BABIP is a stat that measures luck." That is a very misleading, if not completely inaccurate, statement. Pitchers have more control over BABIP than which they are given credit. Using AL averages, the Orioles could be expected to give up 314 hits prior to Sunday's game (based on their batted ball data), but have actually given up 307.

Daniel Cabrera has given up 40 hits. His batted ball data predicts 45 hits allowed. For the month of May Cabrera has given up 9 hits compared to a predicted total of 12 hits allowed.

Also, dkdc's use of his career BABIP does make a lot of sense. See below.

SOURCE

Ron Shandler's Baseball Forecaster clearly states this.

[Lt Melmo]

BABIP is a quirky stat because it lacks consistency and has too many variables that can affect it that are out of a pitchers control. With that said take into the consideration the 71 O's and their 4 - 20 game winners

Jim Palmer .251 BABIP

Dave McNally .231 BABIP

Mike Cuellar .239 BABIP

Pat Dobson .264 BABIP

The O's had 4 gold glovers that year (Brooks, Belanger, Davey Johnson, and Blair). Do you think it was a coincidence the 4 starters had low BABIPs?

Take a closer look at Pat Dobson's career BABIP. In 1970 with the Padres he had a BABIP of .305. Then he posts .269 and .258 in 1971 and 1972, respectively. His BABIP jumps back to a normal level of .298 when he was traded to Atlanta after the 1972 season.

I don't think anyone's going to deny that defense has a lot to do with BABIP. An attentive analyst will use the pitchers' team BABIP as the standard for regression rather than the one-size-fits-all range of .270-.320.

[Remember The Alomar]

Well...it IS more democratic.

Goddamn fascist K's....

What a great film.

[hoosiers]

Just because you don't understand it, and it is pretty obvious that you don't, that doesn't make it pure bunk.

I understand BABIP just fine, thanks.

The stat says little that is conclusive, IMO, such that reliable analysis can be built upon.

[Hallas]

The fact is that there is a lot of noise.

By the way, when you see me making judgments on a pitcher based on BABIP, let me know.

I wasn't really referring to you with the judgements. I just hit quote because you mentioned that you need multiple seasons of data to filter the noise out.

[Lucky Jim]

I understand BABIP just fine, thanks.

The stat says little that is conclusive, IMO, such that reliable analysis can be built upon.

Old #5 Fan...is that you?

[Baltimoron]

One fun thing we know is that extreme ground ball pitchers tend to induce weaker groundballs, and thus do better on the BABIP of their groundballs than less groundball prone pitchers, all else equal.

---------------------------------

From the available pitchf/x data we know DCab has had much more movement on fastball this year than last (Go here), and he is also using it more (almost 84% of the time).

DCab's fastball has been tremendous this year, easily one of the best in baseball. DCab is also throwing his breaking pitch much tighter (notably with much less vertical break, allowing him to throw it for strikes more often) and less often (about half as much as lest year), and has worked in a change about 4% of the time.

Perhaps not coincidentally (or perhaps), he has given up more groundballs and fewer line drives.

------------------------------

There is obviously an enormous sample size issue, and his BABIP is quite unlikely to remain at .226* or so, but it is possible he may be a better pitcher and one who we can expect to be better than his historical BABIP skill.

Anyway, the big curve was more of an out pitch, and in eschewing it almost entirely for more fastballs and a more slider like curveball, it appears DCab is much more focused on attacking the strike zone (although he may still be a little wild) and challenging hitters as opposed to trying to "strike them out".

DCab is throwing more strikes, many more first pitch strikes and fewer pitches per at bat than in the last few years. Hit show hide pitch data summary

---------------------------

* Its also worth mentioning the Os defense has been perhaps a little above average too, which also plays into DCab's BABIP

[Baltimoron]

From the catcher's perspective. Vertical is how much an unspun ball would drop in a vacuum.

I can see much better horizontal (i.e tailing) movement on Dcab's fastball and a little less rise. Far fewer "straight" fastballs and basically none of those ugly looking flatish "cut" fastballs.

And the breaking ball is tighter, with much less vertical drop.

DCab's 2007 pitches

<img src="http://baseball.bornbybits.com/plots/gifs/Daniel_Cabrera.gif">

DCab's 2008 pitches

<img src="http://baseball.bornbybits.com/2008/gifs/Daniel_Cabrera.gif">

(Keep in mind the labeling system gets better the more pitches as the season goes along and the data, while not perfect, is fairly insightful). The labeling of the 2007 breaking ball as a slider as opposed to curve is more semantics, as DCab's 2007 pitch had lots of downward break, but also has a lot more horizontal than a true 16-6 curve. Either way, 2008's breaking ball looks tighter.

[Baltimoron]

If, in fact, BABIP requires more than one season's worth of data to become stastically significant, I have a significant problem with it, as well as any data that requires as much. There are too many variables that can arise when a pitcher goes from one season to another. For example, the pitcher might tinker with his mechanics, causing differences in his pitch velocity and movement, and resulting differences in his pitch f/x data. Now that we have a relatively uniform method for tracking a pitcher's movement with his pitches, I think it will be come easier to calculate a pitcher's expected BABIP based on his pitch movement going forward. However without that bit of information I feel like there's too much noise in general to make any kind of judgements based on BABIP.

Not to argue that you position is per se unreasonable, but I trust you are consistent and thus don't ever cite to any stat, offense or pitching or defense, as pretty much everything takes a season or many more to stabilize. Something to think about.

I think many batters tinker and change there approach too. For an extreme example, I recall Cal seemingly always with a new stance year to year.

I agree about the pitchf/x stuff, but I doubt we will get to predicting BABIP. But I think pitchf/x data can be helpful in our attempts to differentiate between what might be observed and expected variance as opposed to an underlying change in a player's ability.

BABIP is kinda like world class sprinting. Nobody really wins in world class sprinting because they are "faster"* than everybody else. Everybody is really fast. Even the pretty good athletes dropped out long ago. They differentiate themselves by good starts and have great form and condition well and accelerate through according to plan.

Its sorta like that with MLB pitchers. The guys who get hit hard aren't even out there anymore. Most MLB pitchers aren't that different in BABIP prevention, and even the best one's aren't enormously better. Much more important in general are Ks, no walks and getting groundballs. But the ability to avoid hard contact a little better than your peers certainly helps too.

*fast in the actual running sense, like which kid on the yard is fastest sense, obviously form and start go into how fast you run.

[DrungoHazewood]

I've posted before, this BABIP stuff is crazy math trying to normalize a pitcher's balls in play versus other pitchers. It means nothing to me.

The statistic is born of the thought that pitchers have little say regarding the chances of a baseball put into the field of play. I think it's pure bunk.

Just saying a recording of some aspect of performance is pure bunk tells us a lot about your perspective. No statistic is pure bunk, they all have some connection to reality. RBI tells us something, pitcher wins tell us something. Batting average isn't worthless. BABIP certainly tells us something useful. One might disagree exactly what that is, and how much of that is the result of the pitcher's skills, but you're being way too dismissive.

[Hallas]

Not to argue that you position is per se unreasonable, but I trust you are consistent and thus don't ever cite to any stat, offense or pitching or defense, as pretty much everything takes a season or many more to stabilize. Something to think about.

I think many batters tinker and change there approach too. For an extreme example, I recall Cal seemingly always with a new stance year to year.

I agree about the pitchf/x stuff, but I doubt we will get to predicting BABIP. But I think pitchf/x data can be helpful in our attempts to differentiate between what might be observed and expected variance as opposed to an underlying change in a player's ability.

BABIP is kinda like world class sprinting. Nobody really wins in world class sprinting because they are "faster"* than everybody else. Everybody is really fast. Even the pretty good athletes dropped out long ago. They differentiate themselves by good starts and have great form and condition well and accelerate through according to plan.

Its sorta like that with MLB pitchers. The guys who get hit hard aren't even out there anymore. Most MLB pitchers aren't that different in BABIP prevention, and even the best one's aren't enormously better. Much more important in general are Ks, no walks and getting groundballs. But the ability to avoid hard contact a little better than your peers certainly helps too.

*fast in the actual running sense, like which kid on the yard is fastest sense, obviously form and start go into how fast you run.

Well, I'm generally much more hesitant to try and predict future performance with pitchers using only past performance. I think that, with hitters, you can at least make educated guesses (and the correlation coefficients with predictive tools would agree with me.)

I try and evaluate pitchers based on their results, and then look at things that they have day-to-day control over (like the stats that 1970 mentioned about strike% and K-rate.) Pitch f/x data is another great thing in this regard and can give us a (more or less) objective picture of what a scout sees when he says "pitcher x has a heavy fastball," or "pitcher y has a sharp-breaking curve." And as I said, we can say definitively that pitcher x did a great job yesterday. We might even be able to answer the "why" to that question (Why did that pitcher succeed? Well according to his pitch data his curveball falls off the frickin' table. Why did the other pitcher fail? he throws too many balls and doesn't strike anyone out.) But I'm pretty confident in saying that our educated guesses based on said information are, well, somewhat uneducated.

With BABIP we basically know that a pitcher has some control over it, but due to a multitude of variables, including defensive skill and sampling errors, it is muddled with noise that make it exceedingly difficult to draw any conclusions from it except in cases where an extreme discrepancy exists. And since pitchers don't follow a normal bell curve when it comes to aging/performance, trying to use multi-year data with BABIP seems to create just as many issues as it solves.

[Frobby]

I wasn't enamored with the initial post in this thread, but the subsequent discussion has been very illuminating. Thanks to all.

[Mashed Potatoes]

I understand BABIP just fine, thanks.

The stat says little that is conclusive, IMO, such that reliable analysis can be built upon.

I understand the argument that there's too much variance between batted ball data when comparing one pitcher to another. However, would you at least concede that meaningful conclusions can be drawn about an individual pitcher's batted ball data over time?

For example if Pitcher X has a BABIP between .280-.290 over his 6 year career, and then his BABIP drops 50 points one year, doesn't that tell us something? It may not be luck, it may be that he was traded to a superior defensive team, it may be that he's simply employed a new strategy, but that strategy/defensive change is undoubtedly represented in the data.

[crawdad]

I'd worry about comparing 2007 Pitch f/x data with 2008 data. 2007 was very much a trial run with many errors. The raw data needs to go under correction treatments in order to get information that "looks" right. The main issue with that is that you will magnify issues with precision and accuracy.

[dkdc]

What I find interesting is the pitch breakdown by count. Cabrera threw 75% fastballs last year and 84% fastballs so far this year, but most of that increase is concentrated in four counts:

	2007	2008		   Count	Percent	Percent	Diff	   0-0	84.72	89.58	4.86	   [b]0-1	50.50	83.72	33.22[/b]	   0-2	84.62	87.10	2.48	   1-0	85.42	86.42	1.00	   [b]1-1	55.00	83.10	28.10	   1-2	61.54	79.31	17.77[/b]	   2-0	97.22	85.71	-11.51	   2-1	84.78	81.58	-3.20	   [b]2-2	59.09	69.84	10.75	[/b]   3-0	100.00	100.00	0.00	   3-1	95.45	100.00	4.55	   3-2	77.14	80.00	2.86	

Last year, when he got ahead of hitters 0-1, 1-1, 1-2, or 2-2, Cabrera strayed away from his fastball.

I don't have the data to prove this, but I believe that it was often a pitch that he threw for a ball, either as a waste pitch or just because he has poorer command of his secondary pitches.

This year, he's going right after hitters with the fastball in those counts. For the most part, it appears this strategy is working:

	2007	2008Count		OPS	OPSAfter 0-1	0.694	0.420After 1-1	0.742	0.486After 1-2	0.547	0.613After 2-2	0.646	0.500