Control Over BABIP: Not Even Game to Game? Options

[Rudy Pemberton]

But when ESPN.com's DIP% reaches 2.07 (the current number for Haren) or 1.64 and 1.34 (for Clemens in 2005 and 2006), I figure there's a helluva lot of luck in those numbers!

Sure, when guys have ERA's well below 2 or well over 6...it's pretty likely that their "true" ERA is going to be much higher or lower, no? Isn't this similar to a team like the Yankees which was consistently outplaying their Pythag for a long time...when you are winning 95-100 games, it's pretty hard to be underperforming isn't it? Don't know if I'm explaining it that well....

I don't know...when I see that the "normal" line drive percentage is around 20% and a guy like Wang is consistently at 14% for 3+ years, it seems like a real coincidence. He doesn't appear to give up hard hit balls and the more he continues to pitch better than the numbers say he should, the more I think something else may be going on. Now, over time and in his decline phase maybe those numbers even out a bit, but I'm kind of more interested in what can happen during a peak? With the caveat that even 600-800 innings over 5 years may be considered a small sample.

[Eric Van]

So for Bronson Arroyo, increasing his strikeout and walk rates, as well as his G/F, results in an increase in his BABIP. Is this relationship (i.e. all the coefficients are positive) typical?

Different pitchers have different relationships, and it's always possible to make some baseball sense of them.

When looking at any sorts of model equations like this, I like to scale the factors by "typical" ranges. It gives quicker impression of the impact of changes in the factors. es more important than his K rate in terms of its impact on K/BB.

I do the same thing when I want to translate a formula into its real-world meaning, except I'll pull the SD from a wider database (all pitching seasons in a time frame, e.g.). Doing it with the pitcher's own SD is an interesting variation.

[Eric Van]

Hough's career ERA was 3.75. Had he allowed not 1582 earned runs, but 1582 + 142.3, or 1724.3 earned runs in the same number of innings, his career ERA would have been 4.08, or 1.09. I like to think of this as the "Charlie Hough ceiling" -- any DIP% of higher than 1.09 may not be due to luck, but it does suggest that the pitcher in question is doing a better job of controlling his BABIP than the best in history could over his career. And, of course, the higher the DIP%, the more the suspicion of a significant component of luck.

<snip>

And I think Randy Jones is the perfect example of what can happen when a pitcher's BABIP luck runs out: 44-26 in '75-'76, 43-69 over his remaining 6 years.

I don't think his luck evened out; I think hitters got used to his sinker and stopped chasing it out of the zone / pounding it into the ground.

Which is why your Charlie Hough rule is a bit less helpful than it might be. I don't see any reason why a guy couldn't widely exceed that ceiling for a year or even two or three. Keep in mind that getting easy contact is very largely a product of deception; you get a low BABIP when your pitches are less hittable than they look. But deception ordinarily has a short shelf-life.

The interesting thing is that the expectation is no different -- you can expect a low BABIP to regress to the mean because of luck, I can expect it to do the same because of increasing familiarity.

[Rudy Pemberton]

I don't think his luck evened out; I think hitters got used to his sinker and stopped chasing it out of the zone / pounding it into the ground.

Isn't it also quite possible that the sinker stopped sinking like it used to, that he had more trouble throwing it for strikes (which leads to throwing it less, and to throwing more pitches in batters counts which I assume result in better contact). Jones BB rate was 1.6/9 in 75-76, 2.6/9 from 77+.

Has there been any studies done on BABIP by pitch, or by count, or by situation? Correct me if someone has pointed that out before, but I think it could really be interesting. My gut tells me that BABIP on 2-0 counts must be much different than 0-2 counts, but maybe not?

[richard]

Has there been any studies done on BABIP by pitch, or by count, or by situation? Correct me if someone has pointed that out before, but I think it could really be interesting. My gut tells me that BABIP on 2-0 counts must be much different than 0-2 counts, but maybe not?

Weren't the Sox very successful on 3-0 counts in 2004? I seem to remember that swinging on that count was a big thing that year, with very favorable results. Since we haven't seen it happen very much since then, was that approach associated with Papa Jack or the front office?


edit: Bringing this up as a possible data point in favor of your hypothesis.

This post has been edited by richard: Jun 11 2007, 03:49 PM

[JimBoSox9]

2004 Red Sox:
2-0 count
180 AB, 76 H =.422 BABIP
0-2 count
441 AB, 205 K, 80 H = .340 BABIP

2005 Red Sox:
2-0 count
189 AB, 70 H = .370 BABIP
0-2 count
444 AB, 205 K, 72 H = .301 BABIP

2006 Red Sox:
2-0 count
183 AB, 63 H = .344 BABIP
0-2 count
451 AB, 196 K, 62 H = .243 BABIP

Edit: 2004 Red Sox with a 3-0 count, 13 AB, 10 H (.769 BA), 142 BB

This post has been edited by JimBoSox9: Jun 11 2007, 04:22 PM

[richard]

2004 Red Sox:
2-0 count =.422 BABIP
0-2 count = .340 BABIP

2005 Red Sox:
2-0 count = .370 BABIP
0-2 count = .301 BABIP

2006 Red Sox:
2-0 count = .344 BABIP
0-2 count = .243 BABIP

Seems like the relative BABIPs make sense intuitively in both directions, as one would expect better results in a hitter's count (explaining the differential between 2-0 and 0-2), and the offense has been declining since 2004 (explaining the declining BABIP from year to year). Also, is anyone else amazed at the consistency of the ABs for each count from year to year?

Does it follow, then, that a pitcher's control over BABIP would be based largely on getting ahead in the count? If that's the case, maybe better pitching metrics would be strike/ball ratios and K/BB, both of which should have some correlation but maybe not completely?

[Rudy Pemberton]

That was my theory, and the data supplied is pretty interesting. Last year there were a bunch of guys like Delcarmen and Hansen who appeared to be having bad luck, but I thought it was a lot of getting behind in the count and then getting hammered because they were behind in the count. I don't know, just a thought but would be interesting to keep digging, as well as trying to see the babip in different situations as well as by type of pitch. Might we be on to something?

[JimBoSox9]

That was my theory, and the data supplied is pretty interesting. Last year there were a bunch of guys like Delcarmen and Hansen who appeared to be having bad luck, but I thought it was a lot of getting behind in the count and then getting hammered because they were behind in the count. I don't know, just a thought but would be interesting to keep digging, as well as trying to see the babip in different situations as well as by type of pitch. Might we be on to something?

Craig Hansen 2006, with a 2-0 or 3-1 count:
12 AB, 8 H (.666 BABIP)

This is interesting, I'll probably play with numbers from other teams tonight if I don't get too hammered at the Kells beer pong night. it might shed some light to compare individual pitchers BABIP from certain counts to the league average.

My dream is to be able to evaluate balls in play by velocity and trajectory, but until that data is available we'll have to settle on BABIP.

[OttoC]

I've been thinking about this from a non-mathematical point of view. Take a coach with a fungo bat and ball in hand. He can hit ground balls, pop-ups, liners, flies with exquisite control. When I was in grade school, we played softball and I could place the ball just about where I wished. I'd generally go with the pitch, though. If I wanted to pull the ball, I'd lookm for something inside; if I wanted to hit to right, I'd look for a pitch up and away. I hit line drives almost exclusively but if I wanted, I could hit grounders, flies, etc.

When I got to high school, I got a chance to play baseball and while I still made good contact most of the time, taking a full swing, I did hit my share of grounders, flies, and pop-ups. What I couldn't do against the faster pitchers was control to which field I hit. I might be able to hit to the right-side or the left-side but placing the ball down the line or away from a fielder was more a matter of luck.

In the service, I got to bat against a very, very good fast-pitch hurler, and the only way I do anything against him was to get way up front in the box, take a wide-open stance and slap at the ball (and probably make an out).

The point is that up to a point, I could control where and how I hit the ball. It's like a generalized Peter Principle: "in evolution systems tend to develop up to the limit of their adaptive competence." There are batters who possess the reflexes and hand-to-eye coordination to cope with 98-mph fastballs and have some control over where they hit the ball. Put the coach with the fungo bat in a large wind tunnel with 60-mph swirling winds and see how he does...akin to facing a knuckleballer, perhaps.

[paulftodd]

It makes sense that hitters would have better BABIP when ahead on the count when they can look for a pitch and location they like, as they are more likely to hit the ball harder and to the location they want than when taking defensive swings on pitchers counts.

Tango had looked into the effect of BABIP on count from Tom Tippets data the variation was from .282 to.311 except for 3-0 count which was .343 (SSS)

http://www.insidethebook.com/ee/index.php/...tting_by_count/

BABIP = (H-HR)/(AB-K-HR)

[Timmeh49]

I do the same thing when I want to translate a formula into its real-world meaning, except I'll pull the SD from a wider database (all pitching seasons in a time frame, e.g.). Doing it with the pitcher's own SD is an interesting variation.

FWIW, I think pulling the scales from a wider data set is the better way to go; I was just lazy.

[wade boggs chick...]

2004 Red Sox:

2005 Red Sox:

2006 Red Sox:

Edit: 2004 Red Sox with a 3-0 count, 13 AB, 10 H (.769 BA), 142 BB

I assume that these are Red Sox batting numbers? I would also assume that the BABIP for Red Sox pitchers would be very much like this?

And while I agree that it makes intuitive sense that BABIP will be higher if the pitcher is behind in the count, I'm not sure how it makes sense with the fact that a pitcher's BABIP will not be consistent from year to year. One would think that pitchers have an ability to control whether or not they throw strikes, so a pitcher with a good BABIP would probably be getting ahead in the count and probably has the capacity to do that consistently.

Or maybe it's just that most pitchers really don't have a lot of control over whether they get ahead in the count.

Very interesting.

[JimBoSox9]

I assume that these are Red Sox batting numbers? I would also assume that the BABIP for Red Sox pitchers would be very much like this?

And while I agree that it makes intuitive sense that BABIP will be higher if the pitcher is behind in the count, I'm not sure how it makes sense with the fact that a pitcher's BABIP will not be consistent from year to year. One would think that pitchers have an ability to control whether or not they throw strikes, so a pitcher with a good BABIP would probably be getting ahead in the count and probably has the capacity to do that consistently.

Yes, those were the numbers for Sox hitters, here are the pitcher numbers:

2004 Sox
2-0 count: 129 AB, 37 H = .287 BABIP
0-2 count: 542 AB, 88 H, 227 K = .279 BABIP

2005 Sox
2-0 count: 138 AB, 46 H = .333 BABIP
0-2 count: 573 AB, 105 H, 225 K = .301 BABIP

2006 Sox
2-0 count: 158 AB, 60 H = .380 BABIP
0-2 count: 536 AB, 101 H, 229 K = .328 BABIP

Seems like a less obvious split. As for the BABIP fluctuation from year to year, it would not surprise me at all if the BABIP numbers for "neutral" counts fluctuate a lot more from year to year than the BABIP of counts that heavily favor the pitcher or hitter.

[Rudy Pemberton]

2004 Sox
2-0 count: 129 AB, 37 H = .287 BABIP
0-2 count: 542 AB, 88 H, 227 K = .279 BABIP

2005 Sox
2-0 count: 138 AB, 46 H = .333 BABIP
0-2 count: 573 AB, 105 H, 225 K = .301 BABIP

2006 Sox
2-0 count: 158 AB, 60 H = .380 BABIP
0-2 count: 536 AB, 101 H, 229 K = .328 BABIP

Hmm....I wonder if this is a function of the pitching being so much better in '04 vs. '05 and '06? Guess we'd have to break down by pitcher but I would think that better pitchers make better pitches when down / ahead of the count than lesser pitchers (not a profound revelation, obviously!)

This stuff is really fascinating.

[DamonasaNomad]

I don't think his luck evened out; I think hitters got used to his sinker and stopped chasing it out of the zone / pounding it into the ground.

Which is why your Charlie Hough rule is a bit less helpful than it might be. I don't see any reason why a guy couldn't widely exceed that ceiling for a year or even two or three. Keep in mind that getting easy contact is very largely a product of deception; you get a low BABIP when your pitches are less hittable than they look. But deception ordinarily has a short shelf-life.

The interesting thing is that the expectation is no different -- you can expect a low BABIP to regress to the mean because of luck, I can expect it to do the same because of increasing familiarity.

Exactly. And it's very difficult to tell what's luck and what's skill. When a tennis player hits a topspin lob that lands six inches inside the baseline, he sees it as skill -- after all, it's what he intended to do. If he does it fairly consistently for an entire match, or an entire week, he assumes he's gotten really good at the shot -- until the following week when they mostly land six inches beyond the baseline.

Similarly, when a batter swings and hits either a fly ball or a grounder, the pitcher feels like he did something good. But it still might just be luck that the batter's swing was an eighth of an inch lower or higher than necessary to produce a screaming liner. And a fairly small difference in percentage of those can have a significant impact on a season's ERA. Obviously, sample size, always crucial, has to be particularly large for BABIP to even out. That to me is the most significant aspect of that Charlie Hough 1.09 DIP% number: that no pitcher in more than 90 years of analysis has been able to top 1.10 over a substantial sample size. And whether it's his luck or his skill (or the hitters' approach) that's going to change, Clemens's combined DIP% over the past two years of about 1.50 is due for some serious regression.

I do think it would be nice to know what the best DIP% performances over, say, a consecutive five-year period might have been. Presumably, Tippett's study must have those data.

This post has been edited by DamonasaNomad: Jun 12 2007, 10:19 AM

[Rudy Pemberton]

That to me is the most significant aspect of that Charlie Hough 1.09 DIP% number: that no pitcher in more than 90 years of analysis has been able to top 1.10 over a substantial sample size. And whether it's his luck or his skill (or the hitters' approach) that's going to change, Clemens's combined DIP% over the past two years of about 1.50 is due for some serious regression.

I get this, but a 20 year career is pretty likely to average out in the end, in terms of the highs and lows but it doesn't speak to what a pitcher might be able to do over a peak 3-4 year period, which is somewhat insignficant in terms of big pictures but really significant to the success of a few teams.

[paulftodd]

Interesting article at the THT

http://www.hardballtimes.com/main/article/...-look-at-babip/

Outcomes of in-park balls in play appear random because no ordinary records are kept about how well they were hit. David Gassko’s study detected a predictive relationship between strikeouts per game, home runs per game and BABIP. An explanation for these correlations might be that strikeout and home run rates approximately represent a pitcher’s tendency to allow hard hit balls, a tendency that persists as a skill.

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.

This basically shows what should be intuitively obvious. All BIP are not equal. The harder hit balls are hits 59% of the time, and not hard hit balls are hits 19% of the time. Doing the math, for a BABIP of .300 for all BIP, 25% of the BIP are well hit and the 41% of these balls that are outs are lucky for the pitcher since they were hit at a fielder (and perhaps good defense and positioning), and the 19% hits on balls not considered to be hard hit are bad luck for the pitcher (either due to it was a blooper or seeing eyed GB or just poor defense, or some combination of the 2).

The missing link here is data showing what percentage of balls are well hit or not. We have records for GB, LD and FB rate, but while most LD are well hit, not all of them are, and some FB are well hit but most of those that stay in the park are not. For GB's, the percentage of well hit balls may be significant, but the stats available do not tell us anythingabout them. So we are left with looking at LD rates and GB/FB ratios to explain abnormally high or low BABIP, but these are imperfect. We may know for example that a players BABIP on his GB is below the league average, but since we do not know how hard he is hitting those GB, we can not know if it is bad luck or bad hitting.

I agree with the conclusion in the article that pitchers can control the rate at which BIP are well hit or not, and until we have the stats that record these rates and incorporate them into stats adjusting for a pitchers defense or luck, then the other stats currently being used can be questioned.

This post has been edited by paulftodd: Jun 15 2007, 12:42 AM

[DSG]

Some interesting discussion here. First of all, we've learned a lot about BABIP since Voros originally published his theory six years ago. It is not totally random for sure, but the question is, and always has been how much of that skill can measured through other statistics. I'll try to give as quick a summary as possible of what we have discovered:

(1) BABIP is not random. In fact, one standard deviation of BABIP, according to a seminal paper by Erik Allen and Arvis Hsu, is .009 points, which corresponds to roughly .20 points of ERA. In other words, 68% of all pitchers are affected by no more than +/- .20 runs due to their ability to prevent hits on balls in play, while 95% are within +/- .40 runs. So the effect is there, but not particularly large.

(2) BABIP can be predicted better by a pitcher's peripheral rates and phenotypical information than by his BABIP itself (though his BABIP does add a large dose of predictiveness). Peripheral statistics also correlate with the variation in a pitcher's BABIP from year-to-year (sort of what Eric was talking about). When a pitcher's strikeout and home run rates go up, his BABIP goes down; when his walks or hit batters go up, so does his BABIP.

(3) BABIP is largely correlated with a pitcher's batted ball distribution. Pitchers have tons of control over the number of groundballs and fly balls they allow (and bunts too, though that's really dependent on the pitcher's league and overall ability; it's not like he can force a bunt), but little control over the number of line drives they allow (the evidence is mixed on just how little). Groundballs fall for hits a lot more than outfield fly balls in play, though infield flies (which correlate strongly though not perfectly with outfield flies) are almost always outs. In total, groundball pitchers allow a higher BABIP than fly ball pitchers, but they also seem to allow a lower number of bases per hit on balls in play, thus canceling out the higher BABIP.

(4) The most important determinant of BABIP is line drives, which make up 20% of all balls in play and fall for hits 75% of the time. As I said, pitchers appear to have little ability to prevent line drives, though there is some negative correlation between line drive percentage and groundball rate; in other words, groundball pitchers tend to allow fewer line drives than fly ball pitchers.

(5) Relief pitchers allow lower BABIPs because they can go all-out on the mound. They also have higher strikeout rates and lower ERAs for the same reason. Bad pitchers will allow extraordinarily high BABIPs; the BABIP isn't much of a skill argument really only applies to regular major league pitchers.

For more reading on BABIP, here are some essential links:

Voros' original DIPS article
Keith Woolner's rebuttal to Voros' piece
Erik Allen and Arvis Hsu on the true spread in BABIP talent [PDF]
Clay Davenport on minor league pitchers' control over BABIP [may be subscription only]
Me on how much pitchers control batted balls [you have to buy the book, sorry]
My article about the correlation between peripheral statistics and BABIP

I'll be happy to answer any other good questions that might come up, though I hope I've covered this pretty thoroughly.

[DSG]

And to answer the original question in this topic: Have you ever pitched or bowled or played basketball? Sometimes when you pitch you hit your spots (bowl, hit the pins/shoot, make the basket), and sometimes you don't. Are you a better pitcher (bowler/basketball player) on those days? Not really. By random chance, some days will be really good, some will be pretty good, some average, some poor, some awful. Schilling, who is a really good pitcher as is, had a really good day against Oakland.

It's hard to believe in DIPS when you see that most hits come hard off the bat and could not have been caught, but the point of DIPS is not what happens once the ball is put into play, it's actually what happens before, when the ball is on its way to the mound. The point of DIPS is that if a hitter makes contact on a pitch, but doesn't hit it hard enough to put it into the stands, the chances of that ball being a hit are about the same no matter who is on the mound, as long as it's a major league pitcher.

Most pitches will be tough to get a hit on, but some won't. What DIPS tells us is that major league pitchers have about the same breakdown of those pitches. Over the course of a game, where maybe 30 balls are put into play, that breakdown can of course vary significantly.