No Flashback for Halladay
It feels insane to worry about Roy Halladay. It probably is. For almost his entire career, Doc has been positively machine-like, both in terms of his durability and the caliber of his production. From 2001 to 2011, Halladay logged 2300 innings, an average of 209 per season, and sustained a 148 ERA+ — that is to say, over that entire 10 year period, he remained 48% more productive than the average pitcher when adjusting for park and league. He’s had 7 seasons with at least 200 innings pitched and an ERA+ of 140 or greater; only Christy Mathewson, Lefty Grove, Greg Maddux, Walter Johnson, and Roger Clemens have had more.
When a player builds such a titanic, unwavering track record of reliability, particularly when coupled with a work ethic as intense and well-documented as Halladay’s, the threshold for legitimate concern is pushed well out of sight and mind. We already know, thanks to work like that of Derek Carty, that some of the most important measures of pitcher effectiveness take a very long time to stabilize, much longer than those of hitters. For a pitcher like Doc, the suspension of disbelief when a few indicators go sour in a small sample is even more difficult.
A week ago, we were maybe scouting this new terrain from a distance. In the wake of Halladay’s worst start as a Phillie — 8 earned runs in 5 and 1/3rd innings — it’s worth taking a serious inventory of what we do and don’t know. The early trends that are most troublesome are Halladay’s downturns in velocity and strikeout rate. From 2010-2011, which in terms of ERA+ (166) was the best consecutive two season stretch of his career, Halladay’s sinker averaged 92.3 mph and his cutter clocked in at 90.9. So far in 2012, the sinker has averaged 90.4 mph, and the cutter 88.9 mph.
Both Bill and Paul have written about why this isn’t necessarily anything to worry about. Pitch F/X data can be very misleading if used irresponsibly, and bears the same sample size concerns as any other data. As Mike Fast has pointed out, calibration differences in measurement tools introduce some amount of error into velocity and movement data, and the cooler temperatures of April can depress velocity in any pitcher — it tends to ramp up as the season progresses.
These facts alone should provide a solid foundation for skepticism, particularly with a horse like Halladay. But Halladay’s cutter velocity showed very little variation from month to month in 2010, and only gained about 0.4 mph from April to May and June in 2011, tapering off again as August and September wore on. Similarly, his sinker was right at 92.5 mph in each month in 2010, save a momentary 0.5 mph gain in July, and in 2011 it gained only 1 mph from April to Jun, falling back again in August and thereafter. These velocities did not fluctuate nearly as much as Halladay would need to ramp back up to his usual velocity in 2012, and they were never as low in any single month as they have been this April.
Still, it’s not out of the question that temperature has had some minor effect. Halladay had some unfriendly weather in his first five starts, with an average game time temperature of 55 degrees and no single start exceeding 70 degrees. And in the brutal 87 degree heat on Wednesday in Atlanta, Halladay’s cutter and sinker registered their highest velocities yet. For the sinker, it was a gain of about 1 mph over his 49-55 degree starts, and a gain of about 0.4 mph over his 67 degree start in San Diego. For the cutter, it was a gain of 1 mph over his other starts. Admittedly, though, we’re pretty deep into minuscule sample size tinkering here.
It would be plausible that park equipment-related pitch f/x error could be penalizing Halladay’s velocity, but he’s faced more than twice as many batters on the road (118) as he has at home (55), so while he’s thrown more pitches at Citizen’s Bank Park than any other individual stadium, way too high a proportion of them have been away from home for a mis-calibrated gun at the Bank to be dragging him down. And at any rate, as the below velocity chart illustrates, his home velocities are not out of step with his away velocities.
It would be easier to dismiss these fluctuations in velocity were they not accompanied by a significant dip in strikeout rate. From 2010-2011, Doc struck out 22.8% of the batters he faced. That rate falls a bit short of rotation mates Cole Hamels (23.6%) and Cliff Lee (24%), but it is nonetheless quite good — it’s in the 79th percentile of all pitchers from 2010-2011. Thus far in 2012, that rate has dipped to 16.8%, a drop of a little more than 2 hitters per 9 innings. That kind of decrease can have a big impact on any pitcher’s outcomes. For Halladay, it’s meant that the ball has been put in play by 76% of the batters he’s faced, as opposed to 70% from 2010-2011. In a year like 2011, where he faced 933 batters, that would be 56 more who can appeal to good fortune and the imperfection of fielders instead of the cold determinism of the punch out.
The aforementioned Derek Carty found that strikeouts stabilize for a pitcher when (Batters Faced – Intentional Walks – Hits By Pitch) = 126. For Halladay, that equation comes out to 171. As Carty points out, this does not mean that Halladay’s new strikeout rate is “suddenly […] a perfect representation of his talent.” We have scads of history from which to evaluate Roy’s true talent level. But combined with the velocity dip, the strikeout rate is concerning.
Oddly, though, there was a period of time, six years in fact, when Roy Halladay more closely resembled this current iteration of Halladay than the one we’re used to. Coming to the National League has improved his already-elite outcomes, but I think it’s fair to say that 2008-2011 Roy Halladay is a good summation of what we’ve grown to love about him during his Phillies tenure. In that time, Halladay threw 969.1 innings (an average of 242 per season), posted a 160 ERA+ (!), with a 22% strikeout rate and 3.6% walk rate — both excellent. But this was a sea change from Halladay’s previous profile, the beginning of a new era in his career in which his modus operandi transformed in subtle yet important ways. In the previous six years, from 2002-2007, Halladay had posted 1225.1 innings of 141 ERA+ ball, with a 17% strikeout rate. Then, as in 2012, he was missing significantly fewer bats, and allowing around three quarters of his batters to put the ball in play.
Does 2002-2007 serve as a model for Halladay remaining Halladay with his current strikeout rate and velocity dip? Probably not. ERA estimators thought highly of Doc’s inputs during that period — he had an xFIP of 3.36 and a SIERA of 3.44 (his ERA sat well ahead of both at 3.27). They are decidedly more reserved on his current season, at 3.59 and 3.69 respectively. 2002-2007 Halladay did walk batters at a higher rate, 4.7%, then his 2008-2010 self, but still not quite as high as this season’s 5.8%. This is something that Halladay can fix, even if his velocity and strikeout numbers are not. What really jumps out, though, is his incredible ground ball rate in the 2002-2007 period. At 57.7%, he was consistently one of the best starters in the league at inducing ground balls. When he evolved into the mega-elite Halladay of 2008-2011, this figure dropped to 51.5%. There was a corresponding rise in his BABIP from .286 to .294, but that was just fine; he was allowing many less balls in play with his new strikeout rate.
It’s difficult to pin down why, between these two periods, Halladay dipped his groundball rate and bolstered his strikeout rate to become an even more effective pitcher. It’s tempting to point to an adjustment to his repertoire. If Fangraphs pitch information is to be believed, he dramatically increased the use of his cutter between these two periods, from 9.7% from 2002-2007 to 38.4% from 2008-2011. But, as always, we have to be careful about drawing conclusions from pitch f/x classifications. It’s also possible that my selection of these two periods is an arbitrary chop-up of the sample of data that we have, and that taking accumulated stats from these two chunks is doing a disservice to the analysis. I think, though, that if you look a table of his statistics by season, on Baseball Reference or Fangraphs, the transformation is hard to dismiss.
At any rate, Halladay’s ground ball rate currently sits at 50.4%, so he hasn’t been able to call back to his 2002-2007 self by ramping up the grounders while his strikeout rate dips. The current Halladay does not approach the effectiveness of either of the two periods I’ve detailed, unless he can suddenly get the baseballs hitting the floor a lot more often. We’re back to where we started, hoping that we’re making much ado about a bumpy early season ride from one of the steadiest helmsman in the game. Given the history here, I’m totally comfortable taking a calm, skeptical outlook. I would be lying, though, if I said I wouldn’t be watching the radar gun closely tonight.