In today’s Schrodinger’s Bat I discussed, along with historical infield defense, the new wrinkles introduced in the PITCHf/x system for 2008. Chief among those of course is the classification of pitch types that is now visible in the Gameday client as well as in the XML data that analysts examine. It happens that two clarifications are in order.
First, although I reported that Sportvision is reporting the pitch type and therefore developed the methodology for doing so, that is not correct. In fact, Major League Baseball Advanced Media (MLBAM) is entirely responsible for developing and maintaining the algorithms used to classify pitches. So while Sportvision is creating the raw data, MLBAM is processing it to create pitch types. My apologies to the data wizards at MLBAM.
Second, at the time I wrote the article it was in fact correct to say that “while the algorithm considers some information on other pitches thrown by the pitcher, the subset of pitches it chooses from is not restricted”. However, as of this week I am happy to report that changes have been made to the system to restrict the repertoire for each pitcher, which should in large part solve the problem of a pitch being identified as one that the pitcher doesn’t actually throw. Of course this will likely be a work in progress as profiles are developed for each pitcher, but needless to say it is an excellent next step.