Pitch ID - Cubs f/x vs. Gameday

Taking Gameday as the point of reference, how do my pitch IDs for Jeff Suppan (which include a exhibition game from 3/28/2008) match-up to Gameday's?

Fangraphs and I agree, he has four pitches. Gameday has more.



I find almost all their change-ups, and 100% of curves and sliders. I miss a lot of fastballs and the three "extra" pitches they picked-up.

Using Cubs f/x as the point of reference, how'd Gameday do?



Basically, no difference on curves, I find two extra sliders, which Gameday gives one FS and one FC. Gameday tosses in an FC and two CH where I go FA. Change-ups are a whole different story. I'm ID'ing 14 of them that Gameday doesn't.

Gameday calls two of the 14 sinkers (SI), 3 splitters (?) (FS). and the remaining nine are fastballs (FA).

Here's a chart of the PFX for all fourteen, grouped by Gameday's pitch_type



Interesting. What I think this shows is the difference between matching a single pitch to a model (Gameday) and analyzing large groups to find clusters (Cubs f/x, using, primarily, k-means clusters).