A.J. Burnett of the New York Yankees produced an interesting fastball graph this season.
Note that over the course of this season, his fastball velocity stays very even. Batters, however, swing at it less. Their contract rate remains the same, but they are producing a higher average when they hit the ball.
I have a bit of a background in machine learning. This kind of improved output over time from a learning algorithm would make the programmer very happy. Batters have learned to recognize the fastball when to swing and when not to swing, and that's leading to more balls falling for hits. You can especially see this in counts where A.J. holds the advantage over the hitter. In April and May, batters hit .218 against Burnett's fastball in pitcher's counts with a .251 wOBA. Since then, his BA allowed is up to .318 in those situations with a .286 wOBA.
Burnett's fastball became predictable, and batters are taking advantage.