The Case of the Missing Pitch

I got an email this week telling me that some pitchers in Baseball Mogul 2021 were losing one or more of their pitches in the off-season. For example, if you start in 1998, Jeff Suppan has 5 pitches. But then this happens at the beginning of 2002:

His cutter disappears, his sinker becomes a slider, and his "four-seamer" becomes a generic "fastball". It turns out that this bug is a side effect of the way we collected pitch data over the years.

Before we had access to pitchF/X, we maintained a big hand-edited file with each pitcher's pitch types by year. Much of this was built using information from the Neyer/James Guide To Pitchers. But for the hundreds of pitchers that book omitted (including many who reached the majors after it was published), we scoured sources like Baseball America (and did our best to merge conflicting reports when necessary). This data is all stored in a file called PitchRatings.csv. (You can find it in the "Players" folder inside Baseball Mogul's "Input" directory.)

In 2016, we added a file called PItch_Type_Frequency.csv, with pitchF/X data going back to the 2002 season. Now, instead of hand-editing each new pitcher's repertoire, I could just update one file. And because I viewed actual pitchF/X data as "better" than the hand-edited file, I told Baseball Mogul to overwrite the data in PitchRatings.csv with data from Pitch_Type_Frequency.csv whenever possible. And this is how Jeff Suppan's pitch repertoire got f---ed up between the 2001 and 2002 season.

It turns out that the hand-edited data had some advantages, such as including 37 different pitch types instead of the 7 tracked by pitchF/X. So I fixed this problem in Version 24.11 by telling Baseball Mogul to prioritize data from the hand-edited file when possible — and to not overwrite a pitcher's pitch repertoire in the middle of their career.