I see a lot of analysis of this type going on on the internet. Here's why I have problems with it:

1) There is a strange hypothesized idea about how pitching works

I've seen a number of analyses (like this one) that go like something like this:
"Pitcher A was effective when he threw pitch A this often, so he should just go back to throwing it more often and he'll be successful!"

You cannot make this kind of causal statement without stronger evidence of the mechanisms that drive the pitch mix. Does bad pitch mix cause bad pitching or does bad pitching cause bad pitch mix? To the degree that it's the latter, then there's pretty limited utility in examining pitch mix for prediction purposes. There are so many factors that go in to pitch mix (type of batter faced, game situation, "stuff the pitcher has that day") that these kinds of analyses seem nearly useless. To me, it sounds like, "You'll probably feel better from that cold when you stop coughing," it glosses over when and why certain pitches are called for, and I think don't do a good enough job separating cause and effect.

Also, "pitching" isn't just a specific mix of pitches, thrown with particular breaking/speed characteristics - they're thrown in an order: pitches in an at-bat are not independent events. There are important interactions between them that set up subsequent pitches. A FB can set up a slider, and a changeup can set up a FB. I rarely see one of these analyses address the order, location and effectiveness of pitches thrown in at-bats. The degree to which the style of different pitches thrown doesn't seem to be well-characterized either: does rocking a hitter back-and-forth, pitching outside and then inside, have a big effect? To what extent does a pitcher need to change speeds in order to be effective? And how do these factors effect pitch mix? These factors are rarely addressed or even acknowledged.


2) There is "missing data" that biases observed pitch mix

Starters pitch differently throughout a game. Many times a starter will go first time through the order with a FB and curve or something, adding in other pitches the 2nd or 3rd time through a lineup. If their FB isn't working, then they get taken out of the game before they can change their mix to throw more pitch A. You might be lead to believe that not throwing pitch A was the problem, when in fact if the pitcher it was the pitcher's poor FB that led to him not being able to throw pitch A as much as he would have if he had had an effective FB.

Pitch A may mostly be used to generate a ground ball (or swing and miss), and that situation rarely came up. Pitch A might be an 0-2 chase-type pitch, and the pitcher's lack of control with his other pitches led to him having fewer 0-2 or 1-2 counts.

I had some other stuff, but I really need to stop wasting time on this.

All that said, pitch f/x data is awesome. I just wish the analysis were better.