The Case for Marco Estrada, Starter: Part 1 - What Went Wrong

Last week, I looked at Marco Estrada as a potential relief option and came away unimpressed. Now, we'll look at his potential as the fifth starter in the rotation. At this point, it should be no surprise to anyone reading this that Estrada has some some big ups-and-downs the last couple years. Consider the following table:

2014 is split out to isolate his innings as a starter in the last column; of his 138 innings in 2012, nine were in the bullpen but it's not meaningful to split out. Below is his value according to both FanGraphs (fWAR based on FIP and popups) Baseball-Reference (bWAR, based on RA/9 with adjustments for quality of opposition, team defence, and park). What jumps out?

Estrada's decline is a lot bigger measured by fWAR than by bWAR. This is highly unusual. Normally when a pitcher collapses, it's due in strong part to his BABIP spiking and/or his strand rate plummeting, which causes his run allowed (thus bWAR) to balloon. On the other hand, FIP is just based on strikeouts, walks and home runs, the first two of which are much more stable, so fWAR tends to be a lot less volatile than bWAR.

This was not the case for Estrada. Whereas fWAR saw Estrada as a well above average, frontline pitcher in 2012, bWAR saw him as a backend type to begin with. The evaluations aligned nicely in 2013, but in 2014 bWAR evaluated Estrada as a replacement level starter where fWAR overshot that deep into sub-replacement territory.

In fact, the sheer magnitude of his decline from 2012 to 2014 by fWAR is stunning. The season numbers show a decline of 3.5 wins, but even that is masked by decent relief numbers in 2014. As a starter, the decline is four wins, from 3.2 to -0.8. And even that understates things since he wasn't a full time starter. Extrapolating to 180 innings in each year, Estrada went from +4.2 wins/180 innings to -1.3 wins/180, a swing of 5.5 wins. That's the difference between an ace and a scrub. So who's the real Estrada? And what should be used to measure him?

To drill down and try to reconcile this discrepancy and figure out what can be expected going forward, I'm going to take the two metrics driving these systems (FIP and RA/9) and "bridge" from 2012 to 2014 (starting only) by looking at how their components changed and what was the overall effect. Since fWAR includes an adjustment for popups that is not included in FIP, I'll use an adjusted FIP that includes it. I've also adjusted it so the denominator is based on PA rather than IP so it is truly fielding independent (this actually has little effect for Estrada).

Starting with FIP:

In 2012, Estrada posted an adjusted FIP of 3.10, significantly better than his actual FIP of 3.35 because he induced a ton of popups (about 5.5% of all PA vs. 2.4% league average). By 2014, it was over 2.5 runs higher at 5.57, also better than his actual starting FIP of 5.73).

The first thing to note was the decline was broad-based: Estrada went backwards in all components. His strikeout rate fell from 25.2% to 20.5%, which was responsible for his FIP increasing by 0.43. BB* is walks and hit batsmen, and that rate increased from 4.7% to 8.2%, contributing a 0.39 increase in FIP. Estrada also induced fewer popups, though was the smallest factor adding 0.11 to his FIP.

It should come as no surprise (at least if you read Part 1!) that home runs were a huge problem. But the magnitude is remarkable, adding 1.54 to his FIP, or almost twice everything else. And it's not like Estrada was starting from a small base, as in 2012 he had a HR/FB rate of 10.2%, or right around league average. Basically, if everything else stayed at his 2014 rates, but his home runs went back to their average-ish 2014 rate, he's have a FIP around 4.00, which would be quite good for a 5th starter.

Here is the same chart for RA/9, though it's quite a bit messier:

Bridging RA/9 is a lot more involved. Context is the factors Baseball-Reference adjusts for, though it's a small factor of just -0.07 RA/9, mostly due to the decline in offensive production from 2012 to 2014. As with FIP, there's the effects of strikeouts, walks and HBP. The numbers are slightly different because FIP uses fixed run value for each event, whereas for bridging RA/9 I used the actual run values for events in each season. Overall, there's little difference, with those factors adding 0.81 to Estrada's FIP, compared to 0.86 to his runs allowed.

Next are two factors not included in FIP, the damage done by singles, doubles and triples (the latter two grouped together since there are so few triples). Basically, this is what is the "BABIP effect". Allowing fewer singles in 2014 saved Estrada about 0.30 RA/9 compared to 2012; allowing fewer doubles and triples saved another 0.39 RA/9. In total then, "BABIP" factors saved Estrada a considerable 0.69 RA/9, which lines up with his BABIP declining from .298 to .253. Fly ball pitchers like Estrada can have lower BABIPs, so while the .298 was probably due to come down, the .253 is likely to regress a fair bit the other way.

Interestingly, the net effect of all these factors is just +0.10 RA/9, pretty close to 2012. But once again, the home runs dwarf everything else, adding the same 1.54 to his RA/9 as his FIP. This would take his RA/9 to 5.67 (close to his FIP incidentally), but his actual RA/9 was only 5.05. The difference is in the "other" column, and it basically amounts to the sequencing of events and stranding of runners. In general, this doesn't tend to persist for pitchers. Basically, despite a lot of other things going wrong, Estrada had some luck on his side.

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. Tomorrow we'll look at what extent we should expect these to bounce back, and his popups and BABIP suppression too.