Dontrelle Willis? The Dontrelle Willis?
Per-pitch numbers stabilize quicker than per at-bat numbers, or so goes an untested belief. Tested or not, the sample is larger, as the average at-bat lasts between three and four pitches. So you’d think that per-pitch numbers would stabilize up to four times faster.
Well, maybe the relationship is not so linear. But let’s use per-pitch numbers to take a look at Dontrelle Willis, who struck out ten Rockies last night and suddenly looks like a beast. Yes, that Dontrelle Willis.
The first thing that pops out at you is that he’s got his swinging strikes back. After last night’s game, Willis has an 8.8% swinging strike rate, which is not only above average, but also better than he’s shown since his second season in the league. In fact, since 2007, he hasn’t managed a swinging strike rate above 5.5%, in similar samples. His new 6.6 K/9 would be the third-best of his career, too, but that number was about a strikeout and a half lower just yesterday, so let’s focus on the swinging strikes.
Then again, we know this dance started in a different place. Willis got the yips and lost the plate, famously. After giving up just one walk in eight innings last night, his walk rate is under three for the first time since 2005. Back to the per pitch numbers.
Thanks to the pitches and balls numbers, we can try to check this out. In 2007, his last ‘full’ year, 39.4% of his pitches were balls. This year, 34.6% of his pitches have been balls. Another (slight) tick in his favor. In the years in between, collectively, 44.8% of his pitches failed to find the plate. League average is 36.9% this year.
And now we’re stuck in the small sample size doldrums. Let’s go old-school. It looks like Willis pitches solely from the stretch now and is focusing on finding the plate. Part of that is a release point. A warning! This is going to get subjective. On the left, 2009. On the right, 2011. The wonky look comes from trying get Brooks Baseball and Joe Lefkowitz’s pitch f/x plots to line up to scale. Your question, to answer all by yourself: Are the plots more dense, more close together, on the right?
One thing is for certain, the dots are in a new place. Instead of ranging from two to four inches from the plate on the right side (catcher’s view), his release points now range from one to three inches. Perhaps they are closer together, perhaps not.
This is what happens when you try to make decisions based on small sample sizes. You have a few anchors — in this case his per-pitch swinging strike and ball rates — and then you have some subjective work to do. Do you believe that the Dontrelle Willis has found the plate? If so, you better pick him up in your deeper leagues.
With a phone full of pictures of pitchers' fingers, strange beers, and his two toddler sons, Eno Sarris can be found at the ballpark or a brewery most days. Read him here, writing about the A's or Giants at The Athletic, or about beer at October. Follow him on Twitter @enosarris if you can handle the sandwiches and inanity.
Instead of asking us if the points are more dense, why not calculate the variance of distance from the centroid? And if you want to take a sample-size normalized approach, a hypothesis test.