The Most Under Appreciated Batted Ball Type
Call them pop-ups, pop flies, or infield flies. While these batted balls are one and the same, they are not outfield fly balls despite getting lumped together by many baseball sites and analysts. Like Rodney Dangerfield, they get no respect. Infield fly balls are converted into outs about 99% of the time. In other words, only 1% of all pop-ups become hits. By comparison, roughly 75% of all line drives, 25% of ground balls, and 20% of fly balls result in hits (including home runs). Line drives also have the highest run value, followed by fly balls and ground balls. If pop-ups are routinely turned into outs with no advancement by base runners, then they should be treated more like strikeouts for the purpose of performance analysis than anything else. Unlike line drives, fly balls and ground balls, pop-ups and strikeouts have no (or negative) run value. When it comes to breaking out batted balls, I favor Baseball Prospectus over Fangraphs. My preference is not due to the source (BP uses Gameday/MLB Advanced Media and FG uses Baseball Info Solutions) but rather that the former categorizes pop-ups as a separate batted ball event (POP) whereas the latter includes infield fly balls (IFFB) as a subset of fly balls (FB). (You can read Colin Wyers' article, David Appelman's rebuttal, and a thorough discussion at The Book if you are interested in how this data is collected.) Using BP's custom statistic reports, let's take a look at the four different batted ball types as a percentage of all batted balls for 2009 and 2010. As shown, pop-ups account for approximately 7%-8% of all batted balls. While this rate is a fraction of the other batted ball events, it is worth knowing because pop flies are almost always converted into outs. Batted balls represent about 72% of all plate appearances with walks (9%), hit by pitches (1%), and strikeouts (18%) accounting for the balance. While there is a lot of interesting information in the table above, I would like to focus on POP and SO rates as it seems to me that these "automatic outs" could be combined when analyzing pitchers (and hitters, for that matter). Importantly, inducing infield flies appears to be a repeatable skill, much like strikeouts and ground balls, although perhaps not to the same extent. As shown, SO and POP total about 23.5% of all plate appearances. All else equal, I believe that pitchers with higher POP rates — particularly as a percentage of non-SO and GB — should be preferred over those with lower rates. If nothing else, it is my hope that such pitchers may gain greater respect from those who overlook them now. While I want to like SIERA for many of its innovations, I'm not convinced that "pop-ups represent a potential problem for the pitcher in the future." Pop-up rate was allowed to negatively affect SIERA because it is a symptom of the pitcher throwing the ball that generates an upward trajectory, which could lead to an increase in home runs. A pitcher’s skills are throwing strikes, making hitters miss, and throwing with angles and spins such that the trajectory of the ball is downward when it hits the bat. A popup almost always represents an out, but it also represents a potential problem for the pitcher in the future. Moving forward, here are the 2009 rankings of all pitchers with 100 or more innings with an above-average SO + POP rates (SO plus POP divided by PA).
Of these pitchers, Jered Weaver (15.5%), Scott Baker (14.8%), Tim Wakefield (14.1%), Johan Santana (14.0%), David Hernandez (13.3%), Clayton Kershaw (12.9%), Micah Owings (11.6%), Rich Harden (11.4%), David Huff (11.1%), and Todd Wellemeyer (11.1%) induced the greatest number of pop-ups as a percentage of batted balls. Weaver (11.2%), Baker (11.0%), Wakefield (10.8%), Santana (10.1%), Hernandez (10.0%), Huff (9.1%), Owings (8.7%), Wellemeyer (8.4%), Jamie Moyer (8.3%), and Jeremy Guthrie (8.0%) produced the most infield flies as a percentage of plate appearances. Importantly, the rankings of pitchers by SO + POP and POP rates are not meant to identify the most valuable pitchers as neither takes into consideration BB, HBP, or HR rates. However, I wonder if Fielding Independent Pitching (FIP) couldn't be improved by combining SO and POP in its formula, which is typically defined as (HR*13+(BB+HBP-IBB)*3-K*2)/IP plus a league-specific factor (usually around 3.2) to create an equivalent ERA number. The formula for FIP would need to be tinkered to account for the effect of POP as simply adding POP to SO wouldn't work. The multipliers or the league-specific factor would need to be changed to equate the newly constructed FIP with ERA. Here are the top ten leaders for 2010 (among pitchers with 40 or more IP):
Tim Lincecum, Kershaw, Jered Weaver, and Justin Verlander are the only pitchers who ranked in the top ten in 2009 and 2010. Tommy Hanson (14th in 2009 and 5th in 2010), Yovani Gallardo (13th and 8th), and Jonathan Sanchez (12th and 10th) rank in the top 15 both years. The greatest influence on SO + POP is clearly due to the former, yet the latter exerts value on the margin. The ability to induce pop-ups should not be dismissed when evaluating pitchers. Furthermore, it is my belief that certain pitchers have a knack for allowing fewer home runs as a percentage of outfield fly balls than the league average. Saying a pitcher is "lucky" because he has a lower HR/FB rate than the league average is simplistic, as is resorting to xFIP as a standalone measure (especially when a pitcher has a sufficiently large sample size to evaluate). By the same token, labeling a pitcher with a below-average BABIP "lucky" may not be totally accurate either. The analytical community has come a long way on batted ball info. Paying more attention to pop-ups would be instructive in my opinion. Digging deeper into pitcher-batter results as they relate to pitch types, pitch sequencing, ball-strike counts, and bases occupied could lead us to solve some of the mysteries previously ascribed to luck and randomness. For example, pitchers with "plus" changeups may induce more than their fair share of pop-ups and lazy fly balls. More than anything, I hope this article leads to additional discussion and research with respect to analyzing pitchers. * * * Update: Tom Tango sent me an email with a link to Tango's Lab: Batted Ball FIP. He pointed me to posts #8 and #9. Leave it to Tangotiger to have developed a formula for batted ball FIP (bbFIP). The formula is as follows: ERA = 11*[(BB+LD)-(SO+iFB)]/PA + 3*(oFB-GB)/PA + 4.2 Note: the league-specific factor may differ depending on the data source A line drive is like a walk, an infield fly is like a strikeout, and the gap between an outfly and a groundball is about one-fourth the gap between BB and SO. In post #16, Tangotiger lists the results by root mean square error (RMSE) of bbFIP (1.05), SIERA (1.05), and FIP (1.11) and concludes "I’d say that bbFIP is a worthy addition here. Not to mention that it’s in the same spirit as FIP (linear and simple coefficients)." If you have the time and interest, go ahead and read the entire discussion. Brian Cartwright goes into even more detail with numerous tables listing the predictive value of run estimators. As Brian notes, it is important to distinguish between "describing the past vs. predicting the future." I agree. Some skills are more repeatable than others. Guy cautions, "The farther forward you look, the more the skills change/deteriorate." He also warns against "survivor bias" in these studies. Excellent points all. |
Comments
some interesting thoughts. before going any further i would want to see some kind of evidence that a pitcher has some semblance of control over generating pop flies. a glance at those numbers seems to me that it could just be attributed to chance.
if a pitcher can indeed exhibit some degree of control, is it just a byproduct of flyball pitchers or is more the result of pitchers getting beaten and getting jammed on the hands?
Posted by: dutchbrowncoat at May 18, 2010 12:53 PM
Amen. I agree with almost everything you said. I usually like to set baselines for HR/FB rate for each pitcher but these obviously require larger sample sizes than K rates, for example. I would have liked to have seen more tables where you just isolated pop-up % and more talk about the consistency of pop-up % from year-to-year without lumping it in with strikeouts.
Posted by: Dillon at May 18, 2010 12:54 PM
MGL did the heavy lifting on this subject six years ago.
Posted by: Rich Lederer at May 18, 2010 6:17 PM
The question is not so much whether pitchers have "control" over their pop flies, as a function of non-GB BIP, but how much "control" they have, given a particular sample size. Pitchers have "control" over everything. How much control is the important thing. For example, the reason we use FIP or DIPS ERA is not because pitchers have no control over non-HR BIP, but because they have much less control over them than they do BB, SO, and HR. And the reason we use xFIP is not because pitchers have no control over their HR/FB, but because they have much less control over that than they do HR per BIP or per PA.
If you want to lump the pop fly in with K in an FIP or DIPS formula, you better make sure that pitcher control over pop flies is similar to that of the BB, SO, and HR, and greater than that of the other non-HR BIP.
So how much control does a pitcher have to have over an outcome like pop flies, to include it in an FIP or DIPS formula? There is no clear answer. It depends on the sample size of the data. A lot of people do not understand that DIPS and FIP work better on smaller sample sizes and that non-FIP and DIPS formulas like ERC, BaseRuns and even regular old ERA or RA work better than DIPS or FIP with very large samples. In between, take your pick. And when I say, "Works better," I mean in terms of predicting future RA or ERA or in describing a pitcher's true talent.
So before we talk about how important a pitcher's pop fly percentage is, and whether we want to include it in a FIP or DIPS formula, we need to quantify that "control" by looking at something like year to year correlations. My guess is that year to year correlation is not going to be nearly as high as it is for BB, SO, and even HR, and I would hesitate to include it in an FIP or DIPS formula, at least for anything but a small sample of data.
Posted by: MGL at May 18, 2010 10:10 PM
MGL: Thanks. I agree with you on everything, including the fact that ERC, ERA, and RA "work better than DIPS or FIP with very large samples" (despite the reliance by many on FIP and xFIP in such cases).
As it relates to pop flies, I believe such batted balls are, at a minimum, useful in describing the past. They may be less helpful in predicting the future, especially compared to SO and BB. I'm a K-BB-GB (in that order) guy myself as the ability to miss bats trumps all but think there is value on the margin in paying attention to POP/IFFB/iFB as well.
Posted by: Rich Lederer at May 18, 2010 11:07 PM
I agree with Rich that it makes sense to include PU's along with strikeouts. Sure, for predicting future performance we would like to separate the two as strikeouts are more predictive than PU's; however, for retrospective performance the two have essentially equal value and should be viewed similarly.
Posted by: Nick Steiner at May 19, 2010 12:39 AM
Great article. I've been seeing for years discussions about how Barry Zito induces pop outs, and this article clearly validates that thinking.
As long as FIP does not account for pop outs, it will fall down on the job in analyzing pitchers like Zito, who most sabers have been denigrating for years. I wonder how many of the Tom Tippett "Crafty Lefty" category had high SO+POP%.
I'm not really up on the latest and greatest of these, so this is probably a stupid question: I was wondering how Fangraph's tERA would fit into the discussion of comparing SIERA, bbFIP, and FIP.
Lastly, it is interesting that all the Giants main starters were on the Top 41: Lincecum, Sanchez, Zito, Cain. I know Lincecum and Sanchez has the K's to get high on the list, but Zito certainly doesn't and Cain, while good, is not that great at striking out a lot. But I guess both were good at getting pop outs, though not league leading percentages, to boost their ranking for SO+POP%.
Posted by: obsessivegiantscompulsive at May 20, 2010 5:20 PM
tRA is probably the best DIPS metric out there, because it includes all of the stuff as FIP but also a pitcher's batted ball rates. So guys like Zito will be rated more to their abilities.
It's not as predictive as FIP or some of the other stats, but that isn't neccesarily what we are looking for in DIPS..
Posted by: Nick Steiner at May 21, 2010 11:40 PM