Marco Estrada was the surprise of the 2015 season, posting a 3.13 ERA (77 ERA-, or 23% better than league average) despite mediocre peripherals with a 4.40 FIP (107 FIP-, 7% worse than league average).
That differential was almost completely driven by an extremely low .216 BABIP, more than 80 points below the MLB average for starting pitchers. How extreme was that? Of almost 6,600 pitchers seasons of 100 innings between 1970 and 2015, Estrada's 2015 mark was the 9th lowest.
In other words, it was almost inevitable there was going to be a bunch of regression in his BABIP towards league average. Combined with his declining strikeout rate and increasing walk rate, reason to worry about good he'd be.
Instead, his BABIP has been even better in 2016, with a .193 mark that if maintained, would rank as the best on the above list, and the 3rd best since the end of WWII. Estrada is entering truly unprecedented territory as he prepares to come off the disabled list and make his second half debut.
Given that it's still just a season and a half, and that pitcher BABIP is known to have a very high rate of regression towards league average, the natural instinct is to continue to expect a strong degree of correction. We see this in his projections, with both ZiPS and Steamer expecting something in low .270s. But we're now approaching 300 innings of extreme outlying performance, and good Bayesians should always be updating their expectations. So I thought it would be a good time to dig into Estrada's BABIP, and at least understand where the gap is coming from, and then we can have a better idea of what degree of regression should be expected.
Let's start at a high level. In 2015, the league average for starting pitcher BABIP was .297, and in 2016 it's .298. That means Estrada has outperformed by 81 points in 2015, and 105 points in 2016 (90 points for the combined two years). So that's the target to pin down.
Let's start with the most obvious place. Throughout his career, Estrada has been a popup extraordinaire, and popups are basically akin to strikeouts in terms of being automatic outs. Using FanGraphs data, in both 2015 and 2016, the league average popup rate is about 3.2%, or roughly 1-in-30 balls in play (as I wrote this winter, their classification may undercount popups, but we'll stick with these numbers).
In 2015, Estrada had a popup rate of 6.1% (11.6% of his 52% fly balls), or 3% above average. Since this is basically 3% extra automatic outs, right there that accounts for roughly 30 of the 81 points of BABIP outperformance.
In 2016, Estrada's been even more of a popup machine. He's up to 8.6% popups, an incredible 5.4% above average (more than 2.5x the league rate). Again this accounts for a good portion of the outperformance, but what's really interesting and convenient is the unaccounted for portion (105-54) is roughly the same 50 points of BABIP as in 2015.
In other words, the increased BABIP suppression in 2016 is not a fluke, it's basically all more popups. Granted, this is Estrada's highest popup rate ever, so I'm skeptical that he's going to keep up at this rate, so there is likely a little regression coming (his career rate is 6.4% popups). Let's move on to something that doesn't get talked about as much.
On average, about 22% of ground balls go for singles, and about 2% for extra bases for a total of a 24% hit rate or a .240 BABIP. More importantly, these league rates are incredibly consistent across time (2010 to 2015 range is .234 to .241).
What about Estrada? As a Blue Jay, he has a BABIP on ground balls of just .183, or roughly 50-55 points above average. Hey, that's convenient - almost exactly the gap we were looking for. And in 2016, incredibly, he's allowed just 16% singles on ground balls and not a single extra base hit.
What's really interesting about this is that Marco Estrada is an extreme fly ball pitcher, and the research suggest that when it comes to suppressing hits on ground balls, it's higher ground ball rate pitchers that can do that. Or conversely, a pitcher like Estrada should have allowing more hits, not a lot less. So is this the big fluke, where all the regression is going to happen?
Actually, probably not. The other major factor here is infield defense, and the players getting playing time in the infield - Donaldson, Tulo, Barney, Goins - are all very good to elite defenders. It's not just Estrada benefiting from allowing fewer ground ball hits, it's all Blue Jays starting pitchers:
I didn't bother with Dickey since knuckleballers don't have conventional contact results, but the other starters all outforming with a total .206 BABIP, roughly 30 points better than league average. So at least a significant portion of that ground ball BABIP performance has nothing to do with Estrada.
There's another indicator that defensive skill is a significant factor, since there was a big change when a defensive liabilitiy in Jose Reyes was shipped out to be replaced by an excellent defender. To the end of July 2015, Estrada has a .228 BABIP on ground balls. From August on? .145.
That said, Estrada may have some modest contact suppression on ground balls. According to Tony Blengino at Fangraphs, Estrada has allowed
more than below average exit speed on ground balls. However, since Statcast data is new, it's hard to know not only how reliable a half season of data is for a pitcher, but the exact relationship between lower exit speed on grounders and BABIP.
When Estrada signed his extension this winter, my conclusion was the success of the deal would hinge on whether Estrada could continue to suppress contact well. So far, so good; and after looking through the data, he probably can sustain a lower BABIP than previously thought since none of the infield is going anywhere anytime soon (except maybe the DL, but there's depth there with four quality defenders covering three positions).
However, the takeaway that while the infield gives Estrada the ability to sustain a lower BABIP, the credit doesn't belong with him. While something like FIP underrates him due to popups, ERA will overrate him since it credits him for the defenders behind him.
Where does that leave Estrada? If we figure 3% more popups than average, that's 30 points of BABIP. Accounting for some regression, maybe another 25 points due to ground balls. That would leave him with an expectation going forward of something in the low .240 range. That's still significantly higher than what's done so far as a Blue Jay, but it's also a lot less regression than one might otherwise expect.