There has been much written about baseball statistics, most notably from pioneers such as Voros McCracken, Bill James, and Ron Shandler. You can read about the history and use of sabermetrics in real and fantasy baseball in our 2008 Draft Guide Chapter 1: Fantasy Baseball 101. I have also included several reference links at the end of this article in case you would like to read the in-depth information. However, this piece is not intended to give you another boring sabermetric review, so relax and enjoy the show. I do intend, however, to demonstrate that the use of BABIP (defined below) is a bunch of BS (I’m sure that you are aware of that acronym). While I'm not the first to critique this theory, and I'm not about to provide any new revelations at the moment (that’s for a later column), I have however read a few recent articles that used the BABIP stat to draw conclusions and make predictions about a few players that I disagreed with. So, I decided to do a little more digging to help you sort out information and stats that have been flooding the fantasy scene.

BABIP Defined (H - HR) / (AB – SO – HR + SF)

One of the most commonly used statistics in baseball research, especially in fantasy baseball, is Batting Average on Balls in Play, or BABIP. According to research by Voros McCracken, BABIP is commonly used as a red flag in sabermetric analysis, as a consistently high or low BABIP is hard to maintain - much more so for pitchers than hitters. Therefore, BABIP has been used by analysts to spot fluky seasons by pitchers (or hitters), and the analysts predicted that players whose BABIPs are extremely high can often be expected to improve in the following season, and those pitchers whose BABIPs are extremely low can often be expected to regress in the following season. In general, advocating analysts construed that a pitcher’s or hitter’s ability to control what happens to a ball hit in play (e.g., a hit or an out) is mostly random and a result of luck. I disagree with that deduction.

Origins - BABIP and Defense Independent Pitching Stats

The BABIP stat is a sub-stat that was part of a parent theory called Defense Independent Pitching Stats, or DIPS. The stat was derived primarily to evaluate PITCHERS and was intended to be used to determine whether a pitcher has any control over whether a batted ball results in a hit or an out after removing team defense from the equation, and also included adjustments in the formula to account for league variations and ballpark effects. The researcher believed that IF a pitcher could affect hits against him, then it would be evident in his career BABIP stats, because the researcher concluded that a pitcher's BABIP fluctuates annually, while his other "skill stats", such as strikeouts and home runs allowed, remained relatively constant. The obvious conclusion was that "luck" had more of an affect on base hits rather than skill.

While I do concede that there is some luck in baseball, I do not believe that the luck factor is more significant than a player’s skills and abilities.

Concerns about BABIP Applications

Some statisticians agree with the theory, while others have demonstrated that some pitcher's skills do indeed have an affect on whether a ball lands for a hit or not. I have three (3) major concerns about BABIP and it's applications to evaluation of pitchers and the use of BABIP as a predictive indicator of future performance:

1) Probably the biggest concern is that based upon my own research there is no direct correlation between a pitcher's fantasy-relevant stats (ERA/WHIP) and their corresponding BABIP. A pitcher's BABIP fluctuates but their ERA's and WHIP's fluctuate less. Further down in this article, I show that, on average, pitchers with low BABIPs, which you would think is a GOOD thing, actually have higher ERA's and lower percentages of quality starts, while pitcher's with a high BABIP (a bad thing) had lower ERA's and better percentages of quality starts. The problem is likely due to the home run and strikeout statistics, which are major components of the BABIP stat. When a pitcher issues many walks, they are throwing more balls out of the strike zone and there are fewer stikes batted into the field of play to result in a hit. That also shows up in my stat table further down, where pitchers with lower BABIP have higher walk rates. If pitchers are walking batters, the batters obviously cannot get a hit. Pitchers with the higher walk rates invariably end up with higher ERA's and WHIP’s.

2) The next big concern is that the home run stat is subtracted from hits in the formula. This is unfortunate because a home run counts as a hit in my book. Pitchers who give up more home runs will also have lower BABIPs and thus higher ERA's, just like the effect of walks. The reason that the HR stat is subtracted in the original formula is because the HR is a type of hit that is not affected by defense, and the BABIP stat was intended to be used by the creator to evaluate defensible hits – or balls batted into the field of play (not over the wall). This brings me to my final concern:

3) One of the major problems, especially today, is that sabermetricians and fantasy analysts often use the BABIP stat for purposes other than it was intended. As I mentioned above, the stat was part of a parent formula to determine a pitcher's ERA that is independent of defense. The BABIP sub-stat of DIPS was not intended to be used as an analysis tool for hitters, especially because batters face difference defenses all the time, while a pitcher has the same defense behind him, more or less. Also, fantasy analysts use BABIP to evaluate pitchers, such as one revered fantasy analyst did for Brian Bannister (explained later below), and they use it to make assumptions about whether a pitcher is good, or lucky - in general or overall. However, BABIP does not take into account walks, and also subtracts home runs and strikeouts, AND, as I stated in #1 above, there is no correlation between ERA and BABIP, and some pitchers who you would agree are very good have BABIPs similar to those you would agree are bad (Pedro Martinez .281 vs. Eric Milton .287).

One commentator of this article questioned by saying "to say the law of averages, or regression to the mean, does not exist is kind of silly, too. Do you think it's [BABIP] a more useful stat in players past their prime than young players who we don't really know what their ceiling is yet?"

My answer to that is yes, I would agree that the stat may be more applicable to players with significant MLB experience who have already reached their ceiling, especially if their current BABIP is very different from their career BABIP. However, all that means is that their current BABIP may, and I stress, may, revert to his career average BABIP, but even that does not tell us anything useful due to the non-correlation between fantasy relevant stats and BABIP. In real baseball, pitchers are often evaluated by agents, management, and fans based upon their ERA, quality starts, and ultimately, wins.

ANDRUW JONES

For example, after Andruw Jones had eight years of MLB experience at the end of his 2003 season (at age 26), he had a career batting average .269, a career BABIP of .289 (at that time), and averaged 33 HR’s per year. He finished the 2003 season with a .277 average, .293 BABIP, and 36 home runs, all above his career averages. He already had eight years experience in the big leagues (six with full-time at-bats) and posted fairly consistent numbers throughout. Entering his prime year at age 27 in 2004, would you would expect his numbers to perhaps increase a little, stay the same, or revert to normal in the upcoming 2004 season? Consider this before you guess - his BA and BABIP fluctuated each year, except from 2001 to 2003 where his BA, BABIP, and HR’s trended upwards, but not significantly more than his career numbers at that time . Here's where it gets FUN:

In his upcoming 2004 season, his BA dropped from .277 to .261 and his HR's decreased by seven, but his BABIP increased by 10 percentage points (.293 to .303). Then in 2005 his BA slightly increased and his HR's jumped up by 22 more, but his BABIP had the biggest decline in his career at .240 (dropped by 63 percentage points). Guess what? last year his BABIP slightly increased from his monster 2005 season (.240 to .242), but his BA dropped to career low numbers (.222) and he had his lowest HR total since 1999.

Andruw Jones  Summary:

2002-2003: BA, HR, BABIP all increased

2003-2004: BA, HR Dropped - BABIP increased

2004-2005: BA, HR Increased - BABIP dropped

Since the biggest change in his stats was in 2005 where his BABIP dropped massively (.303 to .240), he also had a huge HR season while keeping his BA about the same. Entering his 2006 season, if you would have speculated that his BABIP would increase towards "normal" levels, you would have been correct, However, what did that mean to his important stats? Would you speculate that an increase in BABIP would mean a better BA or more HR, or an overall better season? It makes sense that if more of his batted balls land for hits, he should have better surface stats, right? However, when you look at his 2003-2004 stats above, it shows that his increase in BABIP corresponded with a decrease in both BA and HR. What actually happened?

2005-2006: BA, HR Dropped - BABIP increased

2006-2007: BA, HR, BABIP all dropped together.

Perhaps Jones is an odd-ball example, but it is clear from major league average stats over the last three years that a rise or fall in BABIP does not have a synergistic relationship with BA, HR’s, or any other fantasy-relevant stats. If one falls, the other could rise, fall, or stay the same. Fantasy analysts need to look at all of the relevant stats, and BABIP alone is probably the least relevant stat to try to gain any insight or that may be indicative of future performance. These examples further illustrate the contemporary views by scientific baseball scholars that standard historical stats used to judge player performance and value (i.e., BA, HR, RBI, ERA, WHIP, etc.) are inadequate at best.

BJ UPTON

Last year I read a column published on a major sports web site from a renowned fantasy baseball sabermetrician who also runs his own fantasy baseball web site and publishes his own statistical books annually. In that article, the author argued that Justin Upton’s BABIP was abnormally high, meaning “a crash is coming” and he recommended trading him away. At the time of his article, Upton was hitting .339 with 12 HR’s and 13 SB’s (through July 25, 2007, more than half of the season). However, Upton finished the season at .300, 24 HR and 22 SB’s, doubling his HR total in less at-bats for the remainder of the season. Finishing at .300 surely meant that he hit below .300 for the remainder of the year, but nothing near a “crash,” especially in the HR department. In fact, here are Upton’s post-all-star rankings among all second baseman:

Crash? So okay, that really didn’t bother me, much. I also found it a little ironic, however, that the author paid $18 for Mike Lowell in the 2008 Tout Wars-AL draft (7th most dollars for a 3B in draft) when in 2007 Lowell finished with a career-high BABIP of .337. Oops. Wasn’t he supposed to regress towards his career BABIP of .287? [Editor’s updated note: although Lowell missed several games in 2008 with an injured hip, his BABIP did regress to .282 and his BA also dipped from .324 to .274. His 2008 .BABIP of .282 was very close to his 2006 BABIP of .287 but his HR rate and RBI rate was higher in 2008 compared to 2006].

Law of Averages Fallacy

Much of the theories expressed about BABIP and other sabermetric statistics are based in great part on the Law of Averages, which is a belief that outcomes of a random event will "even out" over a large sample. The “law” typically assumes that an unnatural short-term "balance" will occur.  However, I do remember one interested thing I learned in my college genetics class about 25 years ago: If you flip a coin and it comes up “heads” ten times in a row, ya’ll would think that odds are that “tails” has to come up soon, right? Think again. Every single time a coin is flipped, even after 10 or 20 straight times of landing “heads”, there is still only a 50/50 chance that either “heads” or “tails” will result on each flip. The law isn’t really a scientific law but rather a perception that is not based on any scientific calculations. When applied to baseball, the law ignores non-random events such as player growth and development, skills improvement, health, preparation, and opposing players, etc.

If you buy into the fact that hitters (or pitchers) will eventual regress (or increase) towards the mean of a .300 BABIP, whereas a .300 BABIP has been stated as the average benchmark to which all players will gravitate towards (with the law of averages), then I guess the carrer BABIP’s of Wade Boggs (.344), Rod Carew (.359) and Ichiro Suzuki (.357) are just abhorations. Clearly, all hitters are not run-of-the-mill hitters who just get lucky (or unlucky) from time-to-time. Tell that to Ted Williams, whose career batting average of .344 is higher than his career BABIP of .328.

Good hitters know how to find the holes and get good wood on the ball, perhaps hitting more line drives. Squaring up a pitch for a hit (and HR) is a skill, unless you believe that Barry Bonds couldn’t hit his way out of a paper bag with his “below average” career BABIP of .285. Maybe Bonds needs to play one more year to get his BABIP back up to that .300 benchmark – but he’ll need to get more base hits and fewer home runs. – Hey, I didn’t create the formula.

Okay, I get it…

Yes, I understand that when you look at a player’s BABIP, whether a pitcher or a hitter, you are comparing their current BABIP to their own personal career BABIP, not the league average; and that if a player’s current BABIP is significantly higher or lower than their own personal career BABIP, that something must be out of whack. However, other research has shown that BABIP fluctuates regularly and the fluctuation is random, based on luck, rather than skill. I don’t necessarily buy it.

If one believes in personal BABIP’s then of course that would explain the career-year anomalies. I guess it would also mean that advocates of BABIP don’t put much faith in the 27-year-old breakout theory either (although neither do I). Clearly, it is difficult to translate stats, including BABIP, from the minors to the majors, and I guess BABIP would ignore the natural progression, improvement, or growth of players, and that some of them cannot all-of-a-sudden “get it.” Conversely, it would suggest that players like Andruw Jones shouldn’t all-of-a-sudden suck and continue to suck either. BABIP would suggest that career years are mostly luck and have less to do with a player’s improvement, coaching, prime, juiced balls, opponents, personal issues, performance-enhancing drugs, and a dizzying amount of other factors. So, let’s look at one more example – a pitcher, perhaps.

The Brian Bannister Brainteaser

Quoting from an author in a Spring 2008 USA Today Sports Weekly article: “All pitchers, whether they are Cy Young candidates or hurlers who have never seen the underside of a 5.00 ERA, will post a BABIP that approaches .300. Pitchers whose BABIP is significantly different from .300 can be expected to regress. A high BABIP (bad luck) means a pitcher’s performance is expected to improve. A low BABIP (good luck) means a pitcher’s performance is expected to drop off.”

In that article the author discussed Brian Bannister and referred to his 2008 “low .240 BABIP” and mediocre K/9 rate, further stating “he’s a soft-tosser, and a lucky one at that.” Well, at the time of this article his current .240 BABIP corresponds with his 2.42 ERA through his first four starts. Interestingly, last April of 2007, Bannister posted a similar .243 BABIP, but had a strikingly different ERA of 4.91. Also, in September/October of 2007 he had a “low” BABIP of .253 but a balooning ERA of 7.30. In fact, the month (August) with his second worse BABIP (.290) he actually had his second best ERA (2.90). What gives? He had a 2007 season BABIP of .264. His three-year career BABIP average in the majors is .261. After looking at all pitchers three-year averages from 2005-7, there were 17 pitchers with lower BABIP’s than Bannister. Furthermore, there were 50 pitchers with BABIP’s of .275 or lower. That is many pitchers who, according to the author, to eventually approach to .300. But even that doesn’t mean anything about the players’ success or fantasy-relevant pitching stats, such as ERA, WHIP, or quality starts. While Bannister’s K/9 rate was the lowest of all pitchers in his BABIP range, Bannister also had one of the lowest BB/9 rates as well.

I guess in the author’s estimation, if Bannister's BABIP rises towards Pedro Martinez’s career BABIP of .285, then his ERA should get worse, even though Pedro's ERA is much better and he sports a much higher BABIP. They just don't add up. In fact, I would be more concerned if a pitcher had a high BABIP (say, higher than the league average of .300), because by THAT thinking, his BABIP should eventually lower (if you trust the law of averages), and that could mean less strikeouts and more home runs along with less singles. If you are "selling" Bannister because he has a low BABIP, shouldn't that also mean you should be "buying" Derrick Turnbow, who currently has a .478 BABIP? Heck, he has a lower BABIP than Scott Kazmir at the moment.

If “one of the core elements is a pitcher’s dominance over hitters, which we measure using his rate of strikeouts per nine innings (K/9),” is true, then someone please tell me why Chien-Ming Wang managed to post a three-year average ERA of 3.84 and ranked 9th best overall in quality start percentage when he had the 9th worse (lowest) three-year average K/9 rate of 3.84? Okay, he’s the exception, not the rule, but still...

Ground Ball Stats and Other Useless Information

Since I have spent the majority of this column doing nothing but criticism, you might ask what I have to offer in a positive sense. Well, first of all, the notion of not buying into BABIP when making roster or trade decisions is one way to look at it. Many prognosticators, including Professor Sabergeek, fancy the groundball pitchers, although perhaps not as much as the strikeout pitcher (Bill James stated that curve balls get the most strikeouts, not blazing fastballs, by the way). The theory is that groundball pitchers give up fewer home runs and are also adept at getting more double-plays, thus helping their ERA. However, some basic research of the previous three year averages among all pitchers do not necessarily show that groundball pitchers give up less hits either. It is commonly known that ground balls have a higher chance of being a hit than fly balls (but not non-ground ball line drives).

To finish, let’s compare some stats between ground ball and fly ball pitchers and their corresponding batting average against, ERA, and BABIP and see if we can draw some conclusions, or at least make some interesting observations.

Stats shown in blue text are the best, while stat numbers in red text are the worst.

As indicated by the above statistics of three-year averages among all pitchers, groundball pitchers have the better ERA and QS% but worse K/9 rate, BAA and BABIP. Conversely, the exact opposite is true of fly ball pitchers. The other interesting comparison is between ERA and BABIP alone, where pitchers with the lower BABIP have had higher ERA’s. Therefore, if a current pitcher has an abnormal low BABIP, if his BABIP actually does concede to the law of averages over time during a season (which I don’t believe in all the time), it doesn’t necessarily mean that a rising BABIP corresponds with rising ERA or QS%, in fact, the reverse may be true!  When looking at all starters combined, here is their ERA and corresponding BABIPs:

* MLB Average

The stats for all pitchers show the reverse of groundball/fly ball pitchers, whereas overall, among all pitchers, pitchers with lower ERA’s had a lower BABIP, but the numbers are all relatively close to the league averages, which make drawing any conclusions about the relationship between BABIP and ERA invalid. That notion is further illustrated by the fact between one-quarter to one-third of the pitchers have outliers in their respective ranges (e.g., 26% of pitchers with low ERA have BABIP’s above the league average, most notably J. Lackey, A. Harang, K. Escobar, F. Hernandez, E. Bedard, S. Kazmir, C.C. Sabathia, and R. Oswalt, all of which are above the league average in K/9 and below the league average in BB/9). Note: Harang is among the leaders in most home runs allowed.

Concluding Remarks

Far too much over-analysis is spent on a pitcher’s and hitter’s BABIP, including a hitter’s contact rate, especially early in the season. Variables that are likely to have more of an affect on pitchers’ (or hitters’) effectiveness and production include, but are not limited to: monthly splits, home-road splits, ballpark effects, opposition (schedule), teammates, and defense. A pitcher’s “stuff” and skills, and a hitter’s natural skills, have more affect, while “luck” has the least effect. If a pitcher with seemingly less “stuff” appears to be on a hot streak, and perhaps may have figured things out or mastered his craft, it doesn’t necessarily mean that he cannot sustain his success; and analyzing his BABIP, or any other pitcher’s BABIP for that matter, to draw any resemblance of an accurate prediction about his near future production, is not worth getting brain damage over thinking about it.

“…the knowledge of who will improve is vastly more important than the knowledge of who is good. Stats can tell you who is good, but they’re almost 100 percent useless when it comes to who will improve.” Bill James.

Reference Links

http://en.wikipedia.org/wiki/Defense_Independent_Pitching_Statistics

http://www.diamond-mind.com/articles/ipavg2.htm

http://groups.google.com/group/rec.sport.baseball.analysis/msg/b450fe58c05a5a82