July 2, 2012
Resident Fantasy Genius
Does the Knuckleball Discriminate?
Back at the Baseball Prospectus Citi Field event on June 2, the BP crew and our guests had the pleasure of watching R.A. Dickey extend his impressive scoreless streak with nine innings of shutout ball against the St. Louis Cardinals. Sitting with industry friend Craig Glaser of Bloomberg Sports (who you may recall was on the Fantasy Baseball Panel with Eno Sarris and I at the SABR Analytics Conference), we got to talking about Dickey and how much we loved following him. Craig presented one interesting theory of his that I wanted to test out today.
When watching Dickey, Craig noted how, as a Mets fan, he’s really not that much more scared of the opposing team's great players as he is of their average players. We know that the knuckler is a rare and not-completely-understood pitch, and Craig wondered whether the knuckleball has some inherent properties that neutralize batter talent. He wondered whether good hitters are just as susceptible to being fooled by the pitch as poor hitters are. And when you think about it, this makes some sense. After all, the pitch is rather unpredictable in its movement, and it’s not as if batters have a lot of practice in hitting it. Coming up through amateur ball and the minors, hitters rarely see knuckleballs in the way they do fastballs, curves, and the like. While good hitters see these “normal” pitches over and over, adjust to them, and learn how to hit them, such a process doesn’t really take place with the knuckleball.
To study this theory, I started by looking at all knuckleballs thrown between the start of the PITCHf/x era in 2007 and 2011, limited to known knuckleballers Tim Wakefield, Charlie Haeger, Charlie Zink, and, of course, Dickey. This gives us roughly 17,000 knucklers to examine.
Next, we need to separate the “good” batters and the “poor batters.” There are a few ways to do this, any of which will be necessarily subjective in some respect, but I decided to go a very simple route and take the top and bottom 20 percent of batters in terms of seasonal TAv who faced at least one knuckleball that season and who accumulated at least 300 total plate appearances.
From here, we need to measure a batter’s effectiveness against each individual knuckler. To do this, we need to compare the run value of the situation before the pitch is thrown to the run value of the situation that arises after the pitch is thrown. So, for example, a batter who hits a home run on an 0-2 count will be given more credit than a batter who hits a home run on a 3-0 count since the expectation to do something good is lower in a pitcher’s count. Once we do this for every pitch, we can take the average change in the run value for each of our groups (the “good” hitters and the “poor” hitters) and compare them.
Now, onto the results! I’ve tried to put them into a format that is as easily digestible as possible. What I’ve done is converted the difference between “good” and “poor” batters into a Runs Allowed per Nine Innings format (i.e. ERA but including unearned runs too, assuming 15 pitches per inning).