The Myth of BABIP

[SteelSD]

Originally Posted by TheBigLebowski

Forgive me if this is over-simplification, but isn't the point then that pitchers who get hit hard have high BABIP?

No. Line drives aren't necessarily the hardest hit balls. The "quality contact" tie is only that line drives are "hit 'em where they ain't" quality contact.

If BABIP were indeed "myth", we should find hordes of pitchers who can consistently replicate low annual LD% and who could translate those low annual LD percentages into ERA luck via lower BABIP rates. But we don't find them. What we find is that the vast majority of pitchers are constantly bunched together a percentage point or two above or below the norm. This year's average LD% variance for era qualifiers versus the norm is 0.0003.

For pretty much everyone, LD% is a random driver of a random driver (BABIP). Pitchers who receive low LD% rates aren't necessarily producing them via skill set. Nor are pitchers who receive high LD% rates always getting "hit harder" than those who aren't.

BABIP, by itself is a good indicator as to which pitchers may be getting ERA-lucky because of it. And, yes, we should drill down from there. But using BABIP is vital to pitcher analysis and isn't the least bit controllable for the vast VAST majority of pitchers. There will be outliers, of course. But we can identify them, understand why they're outliers, and determine whether or not those outliers possess true skill sets that will allow them to make their BABIP skills matter as it pertains to ERA luck. When looking at the past three seasons, I've been able to find one ERA title qualifier who may have a meaningful skill set (Brandon Webb).

[Falls City Beer]

Originally Posted by SteelSD

No. Line drives aren't necessarily the hardest hit balls. The "quality contact" tie is only that line drives are "hit 'em where they ain't" quality contact.

If BABIP were indeed "myth", we should find hordes of pitchers who can consistently replicate low annual LD% and who could translate those low annual LD percentages into ERA luck via lower BABIP rates. But we don't find them. What we find is that the vast majority of pitchers are constantly bunched together a percentage point or two above or below the norm. This year's average LD% variance for era qualifiers versus the norm is 0.0003.

For pretty much everyone, LD% is a random driver of a random driver (BABIP). Pitchers who receive low LD% rates aren't necessarily producing them via skill set. Nor are pitchers who receive high LD% rates always getting "hit harder" than those who aren't.

BABIP, by itself is a good indicator as to which pitchers may be getting ERA-lucky because of it. And, yes, we should drill down from there. But using BABIP is vital to pitcher analysis and isn't the least bit controllable for the vast VAST majority of pitchers. There will be outliers, of course. But we can identify them, understand why they're outliers, and determine whether or not those outliers possess true skill sets that will allow them to make their BABIP skills matter as it pertains to ERA luck. When looking at the past three seasons, I've been able to find one ERA title qualifier who may have a meaningful skill set (Brandon Webb).

Very interesting stuff. BTW, people say that ORG is going down the tubes when a thread like this exists. Please.

I have to say, this has given me new reason to reassess my thinking about BABIP and DIPS ERA. I think it all boils to what you are saying: HOW it is applied.

But I have to ask again because I think it might have gotten lost in the last thread: to what degree is the efficacy of BABIP affected by sample size? Is it a better indicator for career starters than relievers? Why or why not?

[IslandRed]

Originally Posted by Falls City Beer

But I have to ask again because I think it might have gotten lost in the last thread: to what degree is the efficacy of BABIP affected by sample size? Is it a better indicator for career starters than relievers? Why or why not?

I don't have any numbers on it. But intuitively, in a situation where randomness plays a factor, bigger sample size = lower standard deviation. Which suggests relievers, who face fewer batters, would have a higher range of BABIPs explainable by randomness and would be less likely to regress to the mean over a certain time frame.

[Falls City Beer]

Originally Posted by IslandRed

I don't have any numbers on it. But intuitively, in a situation where randomness plays a factor, bigger sample size = lower standard deviation. Which suggests relievers, who face fewer batters, would have a higher range of BABIPs explainable by randomness and would be less likely to regress to the mean over a certain time frame.

This is where deduction took me as well, but I wasn't sure, by any means.

[RedsManRick]

For the non statistically inclined, here's a very brief primer on distributions and significance. I'm at work and should be working, so pardon my poor organization. For the stats people, please correct me, it's been a few years -- I know things are slightly different with a variable based on a binomial distribution like BABIP.

Given a normal distribution (bell curve) of a variable (season BABIP in this case), to prove that the difference between a given mean observation (let's say .240 BABIP) and the expected mean (.300 BABIP) is significant (that is, not likely random -- usually <5% chance of being random), we have to look at a few things:

1.) The mean of the distribution (.290)
2.) The difference between the observed and the mean (.050)
3.) The number of observations on which the observed mean is based (IP, PA, etc.)
4.) The overall variance of the distribution (aka standard deviation)

The "true" standard deviation of BABIP is roughly .01. Picture a bell shaped curve centered on .3. As a rule, approximately 68% of all observations fall within 1 standard deviation in either directon of the mean (.280-.300). 95% fall within 2 standard deviations (.270-.310) and 99.7% within 3. However, this assumes infinite sample sizes, which we don't have. Because we're dealing with small sample sizes, the variance expands.

I haven't done the math, so I don't know the exact numbers, but a single season of a .230 BABIP isn't necessarily proof of an ability to control BABIP. In fact, given the hundreds of pitchers, we easily expect a number of pitchers to be outside of the likely range. Remember that a single season has incredible variance, it's only when the total number of observations gets quite high that those ranges mentioned earlier are true.

So if a pitcher is able to put up a .240 BABIP over the course of a career, that's really some good evidence. But a single season or even 2 or 3 seasons doesn't give much statistical evidence.

Steel, perhaps it would be useful to show a handful of guys with the supposed ability to control their BABIP and show the boards what their BABIP has done over their careers, perhaps even cross referenced with ERA for illustrative purposes.

[SteelSD]

Originally Posted by RedsManRick

Steel, perhaps it would be useful to show a handful of guys with the supposed ability to control their BABIP and show the boards what their BABIP has done over their careers, perhaps even cross referenced with ERA for illustrative purposes.

Tall order. But let's try Greg Maddux (stats from baseballprospectus.com)

2006: .315 BABIP
2005: .299
2004: .294
2003: .290
2002: .289
2001: .295
2000: .279
1999: .333
1998: .269
1997: .288
1996: .285
1995: .249
1994: .258
1993: .277
1992: .258
1991: .280
1990: .305
1989: .282
1988: .275
1987: .330
1986: .380

AVG: .288
Median: .287

I hope no one minds that I've excluded the 1986 31.0 IP from the rough average. Again, that's a rough average and if anyone wishes to calculate the actual BABIP for his career, be my guest. Average BABIP is around .290.

On the surface, it appears that Greg Maddux may (I repeat, may) have had some modicum of control over his BABIP- particularly from 1992 through 1995. It's no coincidence that those were also Maddux' four lowest ERA seasons. It's also interesting to me that he spent very few seasons above the .300 mark. And yes, between 1988 and 2002, the only season in which Maddux' ERA ranged above 3.50 was in 1999 (3.57, .333 BABIP).

Yet, we also know that Greg Maddux also produced a career DERA (Defense-adjusted ERA) that was higher than his career NRA (Normalized Runs Allowed). This means he played in front of better-than-average defenses consistently throughout his career.

NRA: 3.51
DERA: 3.61

The DERA rate is 1.03 versus NRA. Mutiply that with his career BABIP and you end up with an adjusted rough average BABIP of .297. Take from that what you will.

Could Greg Maddux have had a modicum of control over his BABIP rates at some point? Maybe, and I'd suggest that if such control existed it manifested itself between 1992 and 1995. That's pretty much what I tend to come back to when studying BABIP patterns- if a pitcher does have any control in the first place, I'd suggest that such control is short-term and fleeting.

[gonelong]

BABIP has its place. It should give you a reasonable indication of how a player might fare into the future, with all things remaining equal. That is a powerful tool to have in your tool box.

A couple of points about BABIP that I think causes people concern.

When you take HRs out of the picture, that inherently closes the gap between a guy like Milton and Pedro. Add HR back in, and you will see that Pedro does control his "BABIP" better than the average starter.

Going a step farther, SLGBIP (inclusive of HR) ... I think it would tell you "more", or at least something different.. When a pitcher gives up contact, what is the result? A weak pitcher would give up more HR & 2Bs than an in-his-prime Maddox would, even if their BABIPs were similar. A guy like Milton would and Pedro would be on different ends of the spectrum again.

I guess what I am trying to say is that we already have plenty of measurements to show that Pedro is a better pitcher than Milton. BABIP isn't striving to show that, but rather tell you if you can expect Milton or Pedro to produce better, the same, or worse ERAs going forward if all things remain equal.

GL

[IslandRed]

I think the main implication of the BABIP studies is not about the pitchers themselves, but in shedding a little more light on the age-old question about where pitching stops and defense begins. Once the assertion is made that pitchers have little or no control over their BABIP, it gives us the ability to look at the differences in BABIP on the team level and credit (or blame) the defense.

GL, I don't remember seeing a definitive work on the subject, but I've seen it suggested by many that while overall BABIP is largely out of the pitcher's control, the extra-base hit rate is not.

[REDREAD]

Originally Posted by TheBigLebowski

Our resident stat mongers here at RZ frequently refer to a pitcher's BABIP during discussions about his effectiveness.

Until about 2 weeks ago, I had no idea what BABIP even was.

For those of you in the dark like I was, BABIP = Batting Average of Balls hit In Play.

.

BABIP might be useful for comparing the same guy year to year. If he suddenly pitches better one year without any explanation.

I think it's a lot less useful for comparing different players or comparing one player against the mean. Just look at Marianno Rivera. Just about all his outs come from balls in play because he gets so many people to pop up. I don't know how to look up BABIP, but I bet he beats the average BABIP every year.
A surface comparision of BABIP would tell you Rivera was the "luckiest" guy in history, but obviously, that is not true.

[Cyclone792]

Originally Posted by REDREAD

BABIP might be useful for comparing the same guy year to year. If he suddenly pitches better one year without any explanation.

I think it's a lot less useful for comparing different players or comparing one player against the mean. Just look at Marianno Rivera. Just about all his outs come from balls in play because he gets so many people to pop up. I don't know how to look up BABIP, but I bet he beats the average BABIP every year.
A surface comparision of BABIP would tell you Rivera was the "luckiest" guy in history, but obviously, that is not true.

Mariano Rivera - only the greatest closer of all-time - is a relief pitching example of a BABIP outlier. He's a freak of nature, like a peak Maddux, but then again like Maddux he's also one of the game's greatest pitchers we've ever seen. Rivera goes back to what I theorized earlier about the game's elite - namely Hall of Fame caliber pitchers - who may tend to exhibit a slightly greater control over BABIP than 95 percent of their contemporaries.

Here's Rivera's BABIP breakdown season-by-season:

1995: .303
1996: .293
1997: .309
1998: .247
1999: .223
2000: .258
2001: .279
2002: .265
2003: .299
2004: .285
2005: .254
2006: .267

Career: .275

Now before we suddenly credit Rivera's success to a low BABIP, first let's remember a few things ... 1) Rivera's K/9 rate is a solid 8.01, 2) Rivera's BB/9 rate is a stellar 2.33, and 3) Rivera's HR/9 rate is an almost impossible 0.45. Since 2003, in 286.1 innings, Rivera has allowed a grand total of nine home runs. That's a HR/9 of 0.28.

Don't we all wish the Reds had a pitcher like him ...

Anyhow, back to his BABIP. Rivera has shown a pretty strong level of being above average in BABIP, however, his overall career BABIP still isn't much greater than above average as it sits at .275. If an average BABIP is in the .290-.300 range, then Rivera's only got the average beat by maybe only 15 points. We're looking for guys here that could consistently be 30, 40, 50 points lower than average.

While Rivera's had individual seasons - 1998, 1999, 2000 and 2005 - with some stellar BABIPs, but he's also had some other seasons - 1995, 1996, 1997 and 2003 - with some below average BABIPs.

Like Maddux, I think there's some evidence here that Rivera has been an outlier. Also, as I previously mentioned, like Maddux he's also one of the game's greatest pitchers so being an outlier should be less surprising. Now, as also previously mentioned - and this is what's important here - just because there may be a handful of outliers does not render the data moot. We can look at hundreds and hundreds of pitchers and may only find a small handful of guys like a Rivera or a Maddux (FWIW, Trevor Hoffman may be a slight outlier too, though less extreme than Rivera).

Lastly, and this could be an important point regarding Rivera and Hoffman, it could be more likely for a dominant reliever to be an outlier than a dominant starter due to the nature of relief pitching. Rivera pitches to far fewer batters per game, per season and for his career than a starting pitcher will pitch to. The data for him, and any reliever, has a much smaller sample size and could also contribute to more outliers existing.

[Johnny Footstool]

The thing that sometimes gets lost in the shuffle when discussing BABIP is that it's only about what happens *after* a ball gets put into play. Good pitchers don't necessarily have much influence over balls that get put into play, but they generally DO have the ability to prevent balls from getting into play, i.e. strikeouts.

[RedsManRick]

double post

[RedsManRick]

From BP:
"Batting Average on balls put into play. A pitcher's average on batted balls ending a plate appearance, excluding home runs. Based on the research of Voros McCracken and others, BABIP is mostly a function of a pitcher's defense and luck, rather than persistent skill. Thus, pitchers with abnormally high or low BABIPs are good bets to see their performances regress to the mean. A typical BABIP is about .290."

That Rivera, with one of the toughest to square up pitches in recent history, only has a BABIP 1.5 Standard Deviations from the norm suggests that if there is an ability to control BABIP, it's relatively minor. I would love to see SLGBIP -- though it would be hard to interpret immediately with HR removed from the equation, it would be interesting to use as a proxy for being hit hard -- sort of like LD%.

It would also be interesting to see the distribution of career BABIP of "good" pitchers, say those with a 7+ K/9, <2.5 BB/9, <1.25 HR/9. See if they fall below the league average as a group, or if there are any particular people who were otherwise good pitchers that were "unlucky". There could be an interesting followup analysis based on pitch type, defense, etc.

What site are guys using for seasonal BABIP?

[Cyclone792]

Rick, for BABIP, LD%, FB%, GB% and other assorted goodies, check out ...

http://www.fangraphs.com/

[TC81190]

The only problem with BABIP is it doesn't account for balls that are actually hit well. You may look at a pitcher and his BABIP may be really high and think "Oh, well he's hit unlucky." That might not be so. He may actually get hit hard. Balls may actually get smoked off the guy.

When coupled with things like his Team's defensive effiency and his own SLG against, then it's very telling.