King Felix and Decreased Velocity: Part Two

Last week I discussed how Felix Hernandez’s pitch selection could possibly offset the adverse effects of his velocity loss. I noted that during the 2010 and 2011 seasons, PITCHf/x informed us that Felix started throwing a sinker. This categorized him into a different group of pitchers based on my cluster analysis, and—what do you know—Felix won a Cy Young in 2010.

But pitching statistics are born in a complex environment with so many lurking variables that it’s either difficult or impossible to figure out the cause of certain changes. For instance, between 2009 and 2010, our King did indeed start throwing what PITCHf/x called a sinker quite a bit more often. That’s what put him into the sinker-balling group to begin with. However, something else happened in 2010, as well: his sinker velocity increased by a whopping 3.5mph, making you wonder if it was really a “sinker” at all.

So how I am supposed to tell what led to his improvement in 2010? (Or better put, his sustained awesomeness since Felix was equally good in 2009 by most measures). When multiple variables change simultaneously, it’s impossible to isolate which was the cause. It would be like weighing myself, taking a shit, then weighing myself on a different scale and discovering that I’d lost a pound. Did I lose a pound because I shat, because I used a new scale, both, or neither? The Felix conundrum is analogous to shitting and using different scales. No, really, it is. Was he able to maintain a seemingly unsustainable ERA because A) He threw a “sinker” nearly 35% of the time? B) He threw it 3.5mph faster? C) Both? Or D) because he did something else?

By using PITCHf/x’s large data set of pitchers, I can at least attempt to isolate potential causes of Felix’s improvements and declines in success, as measured by things like ERA and strikeout rates. The goal is to figure out if he can repel velocity loss and decline at a slower pace than most. Below are some of my findings.

We pretty much know that throwing one’s fastball faster is a good thing, but how does it relate to a pitcher’s changeup? When we just focus on four-seam fastballs, we get an intuitive result: a wider gap between the four-seam and changeup velocities corresponds to a better strikeout rate—probably because it’s easier to fool the batter. But for sinkers, that result is the opposite. There is a statistically significant negative relationship indicating that a pitcher can sustain success if his sinker’s velocity sits closer to that of his changeup. Felix’s four-seam velocity was still in the top 25% of all starting pitchers in 2012, so it’s not an issue yet, but a gap between his changeup (89.3mph) and his four-seamer (92.4mph) needs to continue to exist.

When I allowed the model to sort through information about all three pitches—sinkers, four-seamers and changeups—it told me that a 1-mph drop in sinker velocity would have marginal effects on a pitcher’s strikeout rate (K%), but a 1-mph drop in four-seam fastball velocity would lead to an expected drop of 3% in strikeout rate. That 3% is like the difference between Clayton Kershaw and Madison Bumgarner, for reference. This suggests that perhaps a pitcher in the midst of velocity decline won’t feel the adverse effects quite as much when he throws his sinker. Of course, that doesn’t mean that just throwing more sinkers will make a pitcher better. For example, among only the soft-tossing pitchers in the data set, the four-seamer was still a better pitch.

But here’s one last piece of evidence that having a sinker in the arsenal could be a good thing. I created a regression trying to explain a pitcher’s ERA minus using everything from pitch selection to pitch movement to recent velocity loss. I then traded out 10% four-seamers for 10% of each other pitch, one at a time, and I measured the predicted changes to ERA minus. Here they are:



ERA- Effect













*Here, I added 10% to four-seam frequency, and took 2% away from each other pitch.

Pitchers that threw more cutters, sinkers and sliders also seemed to be pitchers that gave up less runs (on the ERA minus scale). Is it because they threw more such pitches? I can’t say that for sure, but there is definitely a correlation in the data set.

It’s important to remember that there is a lot of noise in the data. The pitcher only has so much control over what the batters choose to do, and thus, all of these models showed a high prediction error. I tried to control for all of the interaction that occurs between the pitches in a pitcher’s arsenal, but important skills like pitch sequencing and location are not easily accessible. In the end, there is a suggestion that, amidst velocity loss, having a something like a sinker could be helpful for Felix Hernandez.

Followup: To get more out of the horizontal and vertical movement data, I still need to divide the pitchers up by handedness. That should come in the next installment of “Playing with PITCHf/x Data.”

  • Casey McLain

    “Sustained Awesomeness” giving way to “Sustainable Awesomeness”

    • Matthias_Kullowatz

      Yes, those are some official statistical terms we use in the Biz…

  • maqman

    All pitchers probably have a fragility factor, although I’m not aware of any way to measure that. From what I know and have observed of Felix my eye test says his factor is lower than an average MLB starter. This doesn’t eliminate him from a major medical issue taking a substantial period of time out of his 7 year contract but his added value potential should compensate for such a period. If he continues to experience a gradual season-to-season decline in velocity though I think his arsenal of pitches and accumulated experience will adequately compensate for the loss through the 7 year time period. The fact that he has some offense behind him will also enable Wedgie to pull him earlier without too much fear the bullpen will blow a lead.

    • Matthias_Kullowatz

      It would be interesting to try and study whether ace pitchers pitch significantly less innings due to good run support…

      • M G Vernon

        I think they do simply because a manager doesn’t feel as much pressure to leave them on the mound if he has a decent lead.

      • Matthias_Kullowatz

        It would be pretty easy to do a game-by-game study for any given pitcher looking at game logs. Maybe that’s my Monday project…