Washington Nationals: Scott Olsen Breaks MLB Gameday?
Scott Olsen pitched a 7 inning shutout against the Los Angeles Dodgers on Sunday, April 25. The official MLB gameday tally, (and thus Pitch F/X tally) showed that Olsen threw 99 pitches. The breakdown – 32 fastballs, 10 sliders, and 57 change ups. What? Did Scott Olsen really throw 57 change-ups? It’s not likely. In Bill Ladson’s post game column, Olsen mentioned that “the slider was working really well. We stayed with that.” In the same article, the Dodger’s Casey Blake said “When he throws that little changeup at 85 [mph], it’s not like a typical left-handed changeup or any changeup. It’s almost like a batting-practice fastball that dives a little bit. We let him off the hook. A lot of times, you do that to a pitcher and the guy sticks around. That’s what we did today.” The question was first asked on FederalBaseball – slider or change-up, what did Olsen throw?
The first thing I looked as was the confidence data assigned to each pitch type classification by Pitch F/X. In the game Sunday, Pitch F/X assigned a confidence level of 75% or below to 35 of the 99 pitches thrown by Olsen. Looking at the 2010 season, Pitch F/X has a pitch type confidence rating of 75% or lower only 14% of the time. That in itself is a giant red flag. Next, I plotted the amount of break each pitch exhibited versus the speed that the pitch was thrown. We would expect a slider to be thrown harder than a change up, and we would also expect the slider to show more horizontal break. First, lets look at the pitches identified as sliders.
This plot looks like a plot that you would associate with change ups. 9 of the 10 pitches were thrown between 79 and 81 miles per hour, one pitch was thrown at 82 MPH and all pitches exhibited little horizontal movement.
Next I plotted the 57 pitches identified as change ups.
This chart is somewhat confusing. While Pitch F/X identified all of these pitches as change ups, it looks like most of these pitches are sliders. When you read the plot, you see that a majority of the pitches broke more than 4 inches horizontally, and they were thrown harder (78 MPH-88 MPH) than pitches identified as change ups.
In order to try to identify the source of the confusion, I looked at the release point of each pitch. The thought behind this is that if the release point is substantially different between the types of pitches, the Pitch F/X algorithm would use that as a discriminator. Here’s the plot.
As you can see from the plot, Olsen released the ball from a nearly identical location when throwing both his slider and his change up.
I also looked at the data from Chad Billingsley. He threw 86 pitches – 6 had a Pitch F/X confidence level less than 75%. Pitch F/X identified 22 of these pitches as sliders, 0 as change ups. Billingsley threw his fastball in the low 90s. The pitches identified as sliders were in the 79 MPH-81 MPH range, but had a slider-like 6-12 inches of horizontal movement.
Without getting in to additional boring details, it appears that Pitch F/X was properly calibrated and working at Nats Park. Pitches identified by Pitch F/X for other pitchers seem reasonably correct based on the data. Which leads us back to our original question – what pitch was Scott Olsen throwing Sunday? Unfortunately, there is nothing in the Pitch F/X data to show why the algorithm missed so many of Olsen’s pitches. Has he developed a new pitch, the “Dippy Batting Practice Fastball”? It’s almost as if the algorithm was reversed, but only for Olsen. Hopefully, whatever the pitch was, he keeps throwing it like he did Sunday for the remainder of 2010.