The Case for Marco Estrada, starter

Yesterday we broke down what was responsible for Marco Estrada going backwards from 2012 to 2014. I'd encourage reading that first, but for ease of reference, below is the RA/9 bridge as well as the brief summary before digging into each component:

In summary then, Estrada's collapse had multiple causes. Home runs were certainly the biggest, but walks and strikeouts were important causes too and should not be minimized.

Strikeouts

As noted in Part 1, in 2012 Estrada struck out 25.2% of batters as a starter, compared to only 20.5% in 2014. Striking out 25% of batters is an elite level for a starter, and Estrada was tied for the 5th best rate among pitchers with 100 starting innings. On the other hand, 2014's mark was barely above the NL league average of 19.5% for starters. For his career, Estrada has struck out 22.8% of batters as a starter.

The single most important factor in striking out batters is generating whiffs (avoiding contact), as can be seen in the chart below comparing the two variables for the 128 pitchers with 100 starting innings in 2012:

Estrada is highlighted in red; he's one of the bigger outliers. With a contact rate of 79.5%, slightly better than the MLB average of 80.8%, he would have been predicted to have a 20.2% K-rate. In fact, of the top 10 in K%, Estrada had the worst contact rate by over 1%. In 2013, Estrada's contact rate improved to 76.2%, but his K% fell to 23.1% - almost exactly what the above model would predict. And in 2014, he has the same 79.5% contact rate, but generated only the 20.5% strike out rate. Again, a really strong fit to the model.

In summary, Estrada's elite K% in 2012 was likely fluky. His career 78.3% contact rate would predict a strikeout rate of about 21%, compared to his actual 22.8%. He's been slightly better than average at being in the zone and throwing first pitch strikes, which may explain of some of that gap. Assuming a bit of a hit for moving to  the better hitting league, a reasonable expectation for 2015 is something like 20-22%.

Walks and Hit Batsmen

Estrada is a strong control pitcher, with a career walk rate of 6.5%. As a starter, it's been even lower at 6.1%, compared to an average of about 7.0-7.5%. He's also avoided hitting batters, with just 10 in 541 career innings, roughly half the MLB rate. In 2012, he walked a very stingy 4.7% of batters, and didn't hit anyone (in 129 innings). In 2013 these rates crept upwards but were still very good; in 2014 they jumped to a 7.8% walk rate and 2 HBP in 107 innings.

Between 2012 and 2014, batters reduced their swing rate against Estrada from 48.6% to 45%. However, this was roughly equally distributed between pitches in and out of the zone, so it shouldn't affect his walk rate much. His first pitch strike rate and zone rate were virtually the same, close to league average. I don't see a good reason for his walk rate spiking the way it did - it was probably unsustainably low in 2012, and high in 2014. Considering his career rate, going forward I'd expect something around 7% (including hit batsmen). Like with strikeouts, this would provide some relief to his expected FIP and RA/9, but it's still closer to 2014 than 2012.

Pop-ups

Estrada is extremely proficient at generating pop-ups, both because he's a fly ball pitcher (about 46% vs. league average 34%) and because he gets more pop-ups per fly ball (just under 14% vs. just under 10%). In 2012, that spiked to almost one-in-five fly balls being popped up (the 5th best of 1,334 pitcher seasons in the last 10 years), and Estrada got 5.5% of all batters to pop-out, over twice the 2.4% league rate.

That's why his adjusted FIP of 3.10 was so much lower than his actual FIP of 3.35: combined with his strikeouts, over of 30% batters were automatic outs. In 2014, his pop-up rate fell to 4.2% as only 12.2% of all his fly balls were popped up. Still an excellent rate, but the 2012 number - being such an historical outlier - is very unlikely to be repeated. I would expect something more like the 12% from 2014.

Balls in Play

Going back to the chart at the top, the one saving grace for Estrada from 2012 to 2014 was improvement on balls in play, both preventing singles and extra base hits. Here's a remarkable stat: in 2014, Estrada allowed 27 home runs against only 20 doubles and no triples. Across MLB, there were over twice as many doubles as home runs, so having more home runs is notable. Historically, it's also quite rare: Baseball-Reference's Play Index indicates just 129 times in MLB history for pitcher seasons over 100 innings.

There's a lot of things that could have contributed to this. Maybe Carlos Gomez robbed some doubles. Miller Park, like Rogers Centre, is a launching pad that may have carried some would-be doubles over the fence. Thanks to the magic of MLB.tv and ESPN's Home Run Tracker, we've got data and video for all of his home runs. 11 of the 27 home runs (41%) Estrada gave up were classified "just enough", average is around 27%. There's definitely some screamers that just left the park and in most cases would have otherwise been extra-base hits (the 6/15 Frazier HR is a good example). I'll go into home runs more shortly, but it seems likely to me that regression in home runs will lead to more damage on doubles and triples.

The other driving factor was a fall in BABIP from .298 in 2012 to very low .253 in 2014. As discussed in part 1, a low BABIP is not unusual for fly ball pitchers. Since 2002 when batted distance is available, the 56 pitchers with a FB% of 44% or better have an average BABIP of .280, about 15-20 points better than average. These flyballers induce a lot of more pop-ups than average, which appears to be the main cause. Estrada's .274 career BABIP would appear reasonably sustainable given that historically he has got even more pop-ups than the group.

A final interesting point is that Estrada's line drive has been dropping over the past couple years when starting, from about 20% to 18% to 16%. On a predictive basis, this is likely just noise around his career average of 18.2%, but another reason to think that the change in his BABIP the last couple years had real underpinnings.

Home Runs

In 2014, Estrada's fly ball rate was more extreme than ever, a career high at just under 50%. That certainly wasn't going to help, but what hurt far worse was his HR/FB ratio exploding to 17.3%. For context of 1,720 starting seasons of 100 innings since 2002, this is the 21st worst (three Blue Jays above him: 2007 Burnett, 2012 Alvarez, 2013 Esmil).

I would expect the FB rate to moderate at least a little bit, especially since they seemed to come at the expense of line drives so it may just be a classification issue. I'd expect something like 46-47% fly balls, but if that's a little higher and the line drive stays down that's actually a good trade-off anyway. If he can generate a few more ground balls, great. But this is really at the margin.

So can we count on the HR/FB rate coming back down? The generic (and valid) point would be that this is a very volatile stat that has a strong tendency to regress towards league average. And prior to 2014, Estrada did not have a serious problem with fly balls turning into home runs. His rate of 11.2% was about 1% above league average, but that's easily explained by Miller Park (2% home/road career differential). From the data on Brooks Baseball, I don't see that he suddenly lost his stuff, which would explain a collapse and be a reason for concern in the future.

But there's one last piece of evidence that really seals things for me. I downloaded pitchf/x data from baseballsavant.com of all balls Estrada allowed in the air for 2012, 2013, and 2014. I filtered only balls hit greater than 300 feet, which are the hardest hit balls:

The 300+ column is how many times batters hit balls hit 300 or more feet. In 2012, 14% of batters faced did that, and 43% of the fly balls Estrada gave up went that far. In both 2013 and 2014, it was a very similar ratio. He didn't suddenly start giving up more hard hit balls. The average distance is essentially unchanged, in fact in 2014 it's actually slightly lower. So the hard hit balls didn't become harder hit. Likewise, the standard deviation is similar, so it's not a case of a bunch of shorter ones and bunch of longer ones.

I'll add a caveat that I haven't checked this kind of data for other pitchers, which would be the last nail in the coffin. But I can't imagine that Estrada suddenly became extremely susceptible to giving up home runs while not giving up harder contact overall. Thus, there's no reason not to expect a pile of regression to his HR/FB rate, to something more on the order of 11-12%.

Wrapping Things Up

Putting everything together gives the following rough projection:

Interestingly, that runs per nine projection works out just slightly higher than J.A. Happ put up last year, when he was worth 1.3 WAR in 158 innings and was a cromulent (see comments below) decent backend starter. Estrada's unlikely to be a star or anything close to 2012, but he doesn't have to be to be provide very reasonable value as a reclamation project.