Infield Pop-Ups and FIP
You may have seen Dave’s article on Fangraphs about some new research that suggests adding infield popups and zone contact% could help FIP identify those pitchers who consistently run lower batting averages on balls in play, and thus, those pitchers whom FIP underrates. The Economists’s Dan Rosenheck presented a paper at this weekend’s Sloan Conference on a metric involving infield fly rates and the whiff rate on pitches inside the strike zone that helped predict whether a pitcher would under or over-perform his FIP.
FIP is, axiomatically, fielding independent – it uses *only* strikeouts, walks and homers to measure what a pitcher’s done. That last bit is important – it seeks to measure what a pitcher’s done, absent any contributions from the defense. No one argues that raw home runs rates/home runs allowed are stable enough to make FIP a “true talent” measure, but as they impact runs allowed so much (and because the defense isn’t involved), they’re included in the formula.
Rosenheck’s work argues (and, to be fair, others have made the point as well) that infield pop ups are essentially fielding independent as well. No major league team is “better” or “worse” at turning pop ups into outs – for all intents and purposes, they’re *always* outs. This makes them look a heck of a lot like strikeouts – they’re nearly always outs, and runners usually can’t advance on them. The fact that a pitcher’s infield fly% isn’t as stable as his K% just makes them more akin to home run rates.
The zone contact term is an interesting one. It suggests that getting whiffs in the zone is correlated with inducing poor contact – something that isn’t counterintuitive, but is cool to see empirically. SIERA (a FIP-alternative metric that uses batted ball data) found that high strikeout pitchers also have lower BABIPs, so again the idea isn’t exactly brand new, but its use in a FIP-ish measure avoids some of the batted-ball classification problems that have bedeviled measures that have tried to improve upon FIP by utilizing batted ball data.
So what happens when we apply these insights to the Mariners projected rotation? The first thing you see when you sort by IFFB% at Fangraphs is that both the “best” and “worst” rotation members were rookies in 2012. That’s something of a problem, as Rosenheck’s formula weights prior year data to attempt to isolate true talent. We’re simply not able to do that with someone like Hisashi Iwakuma. We can use minor league data with Erasmo Ramirez, but it’s not clear how well it ports to the majors. All in all, this probably isn’t a huge surprise – the guys who pitched 200 IP cluster around the league average, and the outliers are those with fewer innings.
Still, if you’re an optimistic sort, you’ll note that Erasmo Ramirez posted both an extremely high infield pop rate and an above-average zone-contact rate, two factors that explain why he was able to post such a low BABIP (.243). That’s great – the model shows why Ramirez did what he did, but what does it tell us about what he’s likely to do in 2013? What’s remarkable about Ramirez’s tenure with the M’s in 2012 is the fact that he bested his raw rate stats from the minors in nearly every category. Forget bettering his major-league equivalencies, Ramirez bettered *his own* rate stats once he moved to the majors. His K rate in the majors was easily higher than anything he posted in the high minors. His walk rate was lower than his AAA stats, though slightly above what he posted in AA. And his infield pop rate spiked in the majors (where he led the team), after several unspectacular showings in the Jackson and Tacoma.
Likewise, Hisashi Iwakuma stands out as having both a terrible infield fly rate and a worse-than-average rate of contact in the strike zone. While he induces more swings *out of the zone* and has a contact rate that’s pretty close to league average, he apparently didn’t throw his (great) splitter in the zone very much. As with Ramirez, you can gin up a pretty decent explanation for these results without breaking a sweat: Ramirez’s change-up is an excellent swing-and-miss pitch that also generates weak contact, ergo, his low 2012 BABIP contains some skill as well as luck. Hisashi Iwakuma’s BABIP will head upwards this year, because hitters have demonstrated an ability to put his fastballs in play, while he’s used his splitter as a chase pitch when he’s ahead in the count. The past is prologue, and now we have a better understanding of why.
It’s easy to create a narrative around these one-year (or less than that in Ramirez’s case) results, the fact remains that we don’t know how much of their performance was talent and how much was luck. We could plug their numbers in to Rosenheck’s formula if we had enough data, but we don’t. Thus, we’re left to speculate about how much Erasmo’s change-up improved or how Iwakuma’s results would look if he utilized his pitches differently. That’s not exactly useful, so I’ll just say that the new metric allows us to do what we would’ve done anyway: if you’re optimistic about Erasmo Ramirez, you now have more data supporting your argument, especially if you’re willing and able to make a causal argument that something *changed* between his MiLB tenure and his MLB success later in the year. With Iwakuma, we don’t even have minor league stats to compare to – his ERA was far better than his FIP. Standard DIPS theory would argue that his K:BB ratio portends good things – that his FIP will fall in line with his ERA as his HR rate drops/stabilizes. This new research would suggest that any HR rate improvement may be balanced by BABIP regression.
With Erasmo, you’ve got to deal with the fact that his major league BABIP is so much lower than his minor league BABIP. With Iwakuma, you’ve got to deal with the fact that you’ve got very little performance record to go on, so the potential that you’re developing a theory based on random statistical noise is depressingly high. Both have one thing in common, though: they have one well-above average pitch that distorts the picture you may glean from their overall statistical record. Erasmo’s change-up generated video-game numbers last year – not only did it get swings-and-misses, it got both ground ball contact AND an impressive number of infield pop-ups (considering the limited data we have). Iwakuma’s splitter was great, even if it didn’t generate a single infield pop – extremely high GB rate and whiff rate made up for the lack of pop flies. With Ramirez, the pitch looked so good that it’s easy to take the data and run: that change-up will allow him to ‘beat’ his FIP going forward and that any regression in K:BB can be weathered thanks to IFFB% and a better sense of when to deploy such a weapon. M’s fans will doubtless do the same regarding Iwakuma, arguing that the lack of IFFBs is manageable because he can consistently generate a different form of bad contact – and anyway, his HR% should regress once he learns to ditch his unbelievably homer-prone four-seamer.
So there you go, new sabermetric research has indeterminate impact on the Mariners 2013 rotation. You’re welcome, internet. This is why I get paid in all of these Pets.com stock options. I’m intrigued by Rosenheck’s research, even if we can’t yet determine what it means for this particular team, and at the very least, I’ll have something else to monitor this year. Those who’ve always found FIP reductionist have slightly more to go on, but it’s still true that every tweak to DIPS theory has – to date – played at the margins. Knuckleballers, some lefties, and whatever the hell Matt Cain does – we’re still not exactly subverting the dominant paradigm here (and we won’t without hit f/x, if you ask me). That said, if there’s one thing this research can do, it’s help identify pitchers who may get overlooked, and Erasmo may be a good example of that. But it’s worth pointing out that Erasmo wouldn’t have scored well by this until he got to the majors; if Erasmo’s the poster boy for Rosenheck’s stat, the credit may go to coaches at the MLB level.