PDA

View Full Version : A different look at BABIP (Hardball Times)

RedsManRick
06-13-2007, 12:55 PM

I won't post the article in it's entirety (though it is available for free above). However, I posted in another thread that BABIP has taken on a life of it's own. I think this point deserves it's own thread as we try to analyze whether certain Reds pitchers have just been unlucky (Mike Stanton?) or truly are doing something which makes them get hit harder (Gary Majewski?)

Many of us, stat-heads and non, have started to run with the idea that BABIP is truly random and that any variation is due to luck. We look at a .350 BABIP and say that the pitcher will necessarily improve and say that the guy with a .260 is bound to regress. This simply isn't true. Yes, there are unsustainable extremities. However, that doesn't mean all variation is random.

It's VERY important to understand that unexplained variance can be random, but is not random by definition. Random is a property of the event which is completely free from control. The outcome of a coin flip is random because we cannot exert any control over it. BABIP has an observed mean and variance, but that variance is due only in part to luck. We simply haven't had access to the data which explains the variation. It may be difficult to prove control, particularly with aggregate data which suggests outliers even in a truly random distribution. However, that doesn't mean real, controllable variation doesn't exist.

Below is the basic premise of the article.

If major league pitchers are skillful at controlling whether batters make hard contact, and BABIP becomes less random in a data-rich world that includes batted ball velocities, then it follows that pitchers have more control over batted ball outcomes than previously thought. BABIP is a lucky number only to the extent that hits occur more or less frequently than predicted by the 0.59 and 0.19 hit conversion rates for well-hit and other batted balls. The pitcher controls his rate of well-hit in-park balls in the same manner as he controls strikeouts, walks, and home runs.

flyer85
06-13-2007, 01:05 PM
I have read that GBs turn into hits at a slightly higher percentage than FBs. Thus you can expect higher BABIPs from GB pitchers. LD&#37; also has an effect and that data has also begun to be tracked as well.

VR
06-13-2007, 01:22 PM
BABIP is a great trend indicator, but by no means the end all of a pitcher's success.

If a pitcher 1 is throwing 70 mile an hour straight fastballs, and pitcher 2 is throwing 95 mile an hours cutters with movement....should we expect BABIP to be equal for balls put in play against those two?

Pitching at this level is all about deception. There's a reason HOF pitchers have lower babip's...and it's not luck. It's because their velocity, control, deceptive pitching, or a combination of the three....cause the hitters to not hit the ball squarely, be off balance, or not get a good swing on their pitches on a consistent basis.

The result is weak grounders and lazy fly balls.

Don't get me wrong, I like the stat...but it certainly is not the end-all be-all of evaluating a pitcher's success.

flyer85
06-13-2007, 01:27 PM
BABIP is a great trend indicator, but by no means the end all of a pitcher's success. I have always paid attention to the extremes because that is where the information is most valuable. If you see a very low (<25) or very high(>35) BABIP there is a good chance a regression to the mean is coming.

Cormier's BABIP, K rate and BB/K rate were the three indicators that said "stay very far away" when the Reds dealt for him last year. I did not care what his ERA was at that point, his PERA(projected ERA) was over 4 at that point. He had simply been a product of good luck and limited usage patterns to that point in the season. Instead the Reds traded for him with a complete misunderstanding of what they had just acquired and how to correctly use him.

VR
06-13-2007, 01:28 PM
Nicely stated flyer.

IslandRed
06-13-2007, 01:43 PM
Good posts. I'd read some followups to McCracken's work that said, essentially, there *is* a difference in BABIPs between good and not-so-good pitchers; it's just not as big as you'd expect and can be completely masked by the year-to-year variances attributable to luck. Another pointed out that baseball talent follows a steep curve, and pitchers not capable of putting up a "true" BABIP close to the mean never reach the majors or don't stay there long. Obviously, you couldn't take the average guy off the street and expect him to have a .300 BABIP.

RedEye
06-13-2007, 02:00 PM
I have read that GBs turn into hits at a slightly higher percentage than FBs. Thus you can expect higher BABIPs from GB pitchers. LD% also has an effect and that data has also begun to be tracked as well.

Am I right to think that even though BABIP is higher for GB pitchers, FB pitchers are still less desirable because there is larger possible damage done by the extra-base hits and HR caused by continual flyballs?

RedsManRick
06-13-2007, 02:01 PM
I haven't done much homework on luck statistics, but I'm sure there's one based on BABIP-exBABIP and adjusted for defense & ballpark. I wonder if BP has something along those lines that's easily accessible.

TOBTTReds
06-13-2007, 02:39 PM
Am I right to think that even though BABIP is higher for GB pitchers, FB pitchers are still less desirable because there is larger possible damage done by the extra-base hits and HR caused by continual flyballs?

I think it might depend on the park. Not sure though. There are some great FB pitchers.

PuffyPig
06-13-2007, 02:39 PM
There's a reason HOF pitchers have lower babip's...and it's not luck. It's because their velocity, control, deceptive pitching, or a combination of the three....cause the hitters to not hit the ball squarely, be off balance, or not get a good swing on their pitches on a consistent basis.

The result is weak grounders and lazy fly balls.

Is this true?

Becuase I've always heard that no pitcher has consitently been able to control his BABIP.

If your 2nd sentence is correct (that pitchers with "cheesy stuff" can control their BABIP why forcing hitters to hit weak grond balls and laxy fly balls), then the whole theory of BABIP is flawed.

flyer85
06-13-2007, 02:43 PM
Am I right to think that even though BABIP is higher for GB pitchers, FB pitchers are still less desirable because there is larger possible damage done by the extra-base hits and HR caused by continual flyballs?correct, FBs have a better chance of ending up as an extra base hit. In addition the HR rate is a product of FB rate, generally around 10% of FBs will end up as HRs. I would think that the each park probably has a native HR to FB rate as well.

RedsManRick
06-13-2007, 03:08 PM
Is this true?

Becuase I've always heard that no pitcher has consitently been able to control his BABIP.

If your 2nd sentence is correct (that pitchers with "cheesy stuff" can control their BABIP why forcing hitters to hit weak grond balls and laxy fly balls), then the whole theory of BABIP is flawed.

And this right here is the fallacy that seems to be making the rounds.

Think of BABIP more like batting average with less spread. Ichiro Suzuki hit .372 in 2004 and .303 in 05. Was he really a much better hitter in 2004? Probably not by that much. Look at Adam Dunn. He hit .215 one year and .266 the next. Same deal. It's not that either of them changed their ability significantly.

The real point is that over the course of a huge sample, no major league pitcher can sustain a BABIP that is more than about 30 points lower than the league average. Mariano Rivera, a poster boy for inducing bad contact, has a career BABIP of .276. He's had single seasons as high as .309 and as low as .223.

By comparison, Eric Milton has a career BABIP of .294, with a low of .241 and high of .329. In 2003, when Milton had a .241 BABIP, Rivera was at .299. And that's the bugger. Within a given year, the ranges of two very different pitchers overlaps quite a bit. Over the long haul Rivera was better at allowing fewer hits on balls in play, by about 25 points of batting average. But in any given season's BABIP doesn't tell you very much about that pitcher's true ability to prevent hits on batted balls.

So where the difference between a career batting champ like Tony Gwynn and a bottom rung BA guy like Rob Deer is about 100 points (.230 to .330), and their averages aren't likely to "cross" in most seasons, with pitchers its somewhere around 50 points (.265 to .315), such that it takes a lot more evidence to confirm the ability and the ability itself is less influential than other skills such as avoiding contact all together while keeping the ball in the zone.

mth123
06-13-2007, 08:23 PM
I have always paid attention to the extremes because that is where the information is most valuable. If you see a very low (<25) or very high(>35) BABIP there is a good chance a regression to the mean is coming.

That is kind of my take too with a little twist. I really don't think that a bad pitcher has an upper limit to BABIP. A pitcher who has bad stats with a BABIP of say .375 may very well simply be a bad pitcher.

OTOH a pitcher putting up good stats based on an extremely low BABIP (say less than .240) is probably the beneficiary of luck and great defense.

I also tend to look at the other peripheral stats like K/9, BB/9 and HR/9 in conjunction with BABIP to get my bearings a little. You have to look at the entire picture IMO.

coachw513
06-14-2007, 12:30 AM
thanks for a very interesting article and topic:thumbup:

Patrick Bateman
06-14-2007, 12:37 AM
Over the long haul Rivera was better at allowing fewer hits on balls in play, by about 25 points of batting average. But in any given season's BABIP doesn't tell you very much about that pitcher's true ability to prevent hits on batted balls.

Couldn't this be widely attributable to the fact that Rivera has played on the Yankees his entire life with great defenses supporting him? If he was playing on the Reds his enitre life, I would bet his BAPIP would be approaching .300.

blumj
06-14-2007, 02:32 AM
It's been a while now since I've thought of the Yankees as having a great defense.

Cooper
06-14-2007, 10:50 AM
I haven't read the articke yet --so if this has been mentioned -my bad....i've viewed BABIP as a way to evaluate luck -but it seems to fit better with those pitchers that have good strike out numbers. For example: if a pitcher has a high BABIP and a high K/9inn. then you can assume his BABIP avg. will regress to the mean? Am i making a faulty assumption?

The poster boy for that would be Sam LeCure. He has a good ratios and his K/9inn. rate is high. His BABIP is very high...i would assume that would come down because he has the supporting elements in his game.

IslandRed
06-14-2007, 10:56 AM
It's been a while now since I've thought of the Yankees as having a great defense.

Or even above-average.

D-Man
06-14-2007, 12:08 PM
Pitching at this level is all about deception. There's a reason HOF pitchers have lower babip's...and it's not luck. It's because their velocity, control, deceptive pitching, or a combination of the three....cause the hitters to not hit the ball squarely, be off balance, or not get a good swing on their pitches on a consistent basis.

The result is weak grounders and lazy fly balls.

I think the article provides a pithy summary for *why* we use BABIP: "Strikeout and home run rates serve as rough-and-ready proxies for a latent contact hardness variable that manifests as batted ball velocity. " The phenomenon that you describe would also show up in an increased line drive percentage, aand you could make BABIP adjustments accordingly.

And finally, a team's fielders matter a lot. Pitchers that play behind great defenses are helped out tremendously. For instance, Tom Glavine is a great pitcher, but he was helped by some great ATL defenses over the years. Paraphrasing Bill James: a lot of what is thought to be pitching is in fact defense.

RedsManRick: Thanks for posting the article.

Although I have to say I find the evidence presented to explain Zito's BABIP improvement (i.e., that he developed a slider against lefties) to be specious. If his BABIP improvement from .368 to .247 from 2004-2005 were real and not due to randomness, you would expect to see continued success against lefties in 2006. But his BABIP in 2006 was .297, just where BABIP would predict it to be. So instead, I see randomness mostly at work here.

The Greinke case is more cogent, particularly the fact that he got behind in the count more often. Cyclone had a great post last season where he attributed Arroyo's 2006 success to how he was retiring the first batters in each inning with greater frequency. This makes sense: hitters fare better when they are ahead in the count, or when there are baserunners, or when runners are in scoring position, or when there are fewer outs in the inning. So these are elements that can and should be tracked to provide context within the BABIP enviornment.

RedsManRick
06-14-2007, 12:16 PM
D-Man, I wouldn't take these two examples as proof of anything per se'. Rather, I think they merely serve as good illustrations that BABIP variation is not only a result of luck but of actual changes in pitcher performance. A sample of 3 years is certainly no statistically significant proof of anything. All of those 3 BABIP years could be due to random variation. However, they do serve to illustrate a concept, to lay the groundwork for a larger study.

As I think he rightly points out, what appears as randomness is often due to a lack of data from the thing which actually explains the variance. Very few things are truly random, they merely are outcomes from a system too complex to explain/predict. In this case, when you add certain pitch-by-pitch data, we see a little more clearly how BABIP might be affected. If we were able to accurately capture defensive performance, I'm sure we'd be able to explain it even more. Maybe Zito's defense sucked in that third year, while he was still just as effective. We simply don't have the data to know. What is truly random about BABIP is the combination of functionally independent events of batted ball location, speed, and trajectory, weather, and all the individual variables of defense including the player, his positioning, his reaction, etc. The exact same batted ball will not always result in the same outcome.

Technically, if we could know everything about everything, we would coin flips predictable too. I guess you start to slip in to chaos theory, but if you flip height, rotation axis & speed, the elasticity of the landing surface, air currents, weight distribution in the coin, etc, you could theoretically know how a coin would land. Of course, those data points are so difficult to capture and the system is so complex, that all we can really observe is the distribution of outcomes.

With batted balls, we can and are capturing more and more data and while we will surely gain the ability to explain batted ball outcomes more clearly, there will always be a certain level of randomness. This article just illustrates that we're explaining a little bit more than used to. What's important to me is to understand if Barry Zito's BABIP mean is truly different from Zack Greinke based on the types of batted balls they allow. Right now, the evidence suggests that there IS an ability to allow a lower BABIP. The problem is that, with some exception -- particularly on the low end, we just can't know for sure if any given BABIP observation is a result of that skill or that natural variation.

D-Man
06-14-2007, 02:23 PM
As I think he rightly points out, what appears as randomness is often due to a lack of data from the thing which actually explains the variance.

I don't disagree with your broader point at all--in fact, I agree with it. My comment was focused more narrowly on the argument in the Zito portion of the article. Namely, I had issue with Isley's conclusion that "the slider enhanced Zito’s other pitches by giving hitters something else to think about—especially left-handed ones." That explanation is especially weak when we examine the three-year sample, as opposed to the two-year sample. Isley is making a logical leap in say that the Zito's slider = enhanced other pitches. I just don't see it. I would look for other reasonable "answers" (or "functionally independent events," in your words :thumbup: ) to explain the variance.

Perhaps the conclusion could be qualified to provide a temporal range (i.e., the slide enhanced Zito's other pitches in 2005.) I suppose these kinds of changes in approach have a short-term effect on BABIP--this data would support that kind of conclusion, but certainly not the broader conclusion Isley makes.