clock menu more-arrow no yes mobile

Filed under:

Velocity Drop Shouldn't Cause Concern for Keuchel

Why early season peripherals shouldn't be an issue for Dallas Keuchel

Troy Taormina-USA TODAY Sports

Early returns from the 2016 Astros haven’t been great, especially from the starting rotation. Cy Young winner Dallas Keuchel isn’t a guy Astros fans should be worried about, at least on paper. However, a recent Fangraphs article from Dave Cameron brings up the idea that Keuchel’s early-season peripherals could point to some underlying issues based on velocity and called strikes.

Last season, Keuchel sat comfortably around 90 mph with his fastball/sinker combo. Though velocity will never be his calling card, it’s a generally-held belief (albeit not one totally based on data) that a consistent major league starter needs to sit around 89-90 mph, at minimum, to have success (no news yet on how Jered Weaver’s low-80s fastball will hold up). It may not be a coincidence, then, to see how Keuchel’s velocity changes over the years have led to his success. In 2011-2012 when Keuchel was a mainstay on the Quad-A express between Oklahoma City and Houston, Keuchel sat around 87-88 on his fourseam and a bit lower on the sinker. A few years later, Keuchel starts hitting 90 mph, and he became a Cy Young winner.

That’s not to say a two or three mile per hour jump caused Dallas Keuchel to turn into this decade’s version of Cliff Lee, as other factors (like Brent Strom) definitely come into play. However, Keuchel’s velocity has been closer to those 2011-2012 levels so far this season. Here’s a visual of Keuchel’s velocity on four pitches with timeframes from all of 2015 and his first three individual starts of 2016:

Keuchel’s batted ball data from last season tells us that a certain contact rate shouldn’t matter a ton- Keuchel is a pitch-to-contact guy, but that contacts leads to softer stuff like groundballs which the Astros defense can easily turn into outs. So far in 2016, he’s allowing harder contact in the form of line drives and a decrease in groundball rate. Here’s Keuchel’s batted ball profile from this season and last:

That’s a pretty substantial jump in LD% (although still around league average) and hard-hit rate. Luckily, Keuchel hasn’t actually been hurt by this trend so far. He’s yet to allow a home run, and hitters are only batting .192 against him. Good numbers, but a bit unsustainable as shown by an xFIP of 3.88.

If Keuchel’s velocity stays in the danger zone and he’s tossing fastballs at 88 throughout the year, his contact rates could stay where they are- those increased line drives (and fewer groundballs) will lead to fewer easy outs and more bad contact, which would obviously be a negative trend. Keuchel would really have to rely on his pinpoint control to locate slower fastballs so they don’t get hammered.

On the other and more likely side, Keuchel is still rounding into form with his first three starts. In his last outing, a dominant performance against the Tigers, Keuchel threw his fourseam and slider velocity closer to 2015 levels. On a more general note, pitchers historically throw slower in April- this could be from cooler weather, or guys just figuring out their stuff again after months off from live game action.

Keuchel’s “issues” with called strikes are also briefly worth discussing. In the article, Cameron believes that umpires could be bearing down on Keuchel’s presence in the bottom of the strike zone last season. This application of the Law of Diminishing Returns is definitely worth considering on a macro level for all game trends with so much available data (Cameron suggests that umpires could be tracking Statcast and Pitch F/X data to see what kind of calls they’re missing). It’s unlikely, though, that this has already caught up to Keuchel. He’s actually getting the same rate of strike calls this season (he has 60 already) compared to last season at 18.5%. The idea Keuchel is getting fewer called strikes isn’t evident yet, but like velocity and everything else in baseball, it could be worth monitoring once sample sizes being to normalize in a few weeks.