Saturday, April 20, 2013

What is Most Predictive of a Good Start?

In this post, I will look at all the MLB starts from April 14th and April 18th and test what is most predictive of a successful start. We have already seen that fastball velocity is a great predictor of a good start, so in this post we will look at overall pitch velocity, release points and release point consistency. Here is the spreadsheet containing all of the starts (it doesn't have every starting pitcher, especially because of some rainouts, but the 5 day net gives us a look at a lot of starting pitchers):

Rather than looking at numbers, here we will look at some scatterplots to check the correlation of each number. I also put lines in each graph that represent the expected correlation, if it was perfect.

Here is the scatterplot of pitch velocity and FIP

Obviously overall pitch velocity doesn't only take in account the velocity of each individual pitch, it also takes into account pitch selection. A high overall pitch velocity says that not only do they throw hard, they threw their hard stuff a lot. Here, we don't see a lot of correlation because there were a lot of starters that threw 85-90 and didnt have good starts. The 80-85 MPH guys didn't have the best starts, but they were also less likely to have horrible starts.

In the next graphs, FIP serves as the horizontal axis. Here is the horizontal consistency (remember, the higher number means least consistent) versus FIP

Not much correlation here. There isn't a whole lot of variation overall (an obvious example being Bronson Arroyo who intentionally changes his arm angle throughout the game), and no real reason for why some of them were horrible starts and some of them were good.

What about just average horizontal release point (the higher the number means the further out)? Here, I assumed that there was no real value difference between lefties and righties, so I changed the negatives (righties) to positives.

Again, not a ton of correlation here, or at least not expected correlation. 1 to 2 feet guys could be anywhere, while the very far out guys seemed to be the best. So perhaps reverse correlation.

Does vertical consistency matter more than horizontal consistency?:

Not really. Again, there wasn't much variation (I took out Arroyo here to get a better look).

What abut just vertical release point:

This is actually probable the best correlative set, and it isn't great either.

Certainly one could have used a different stat than FIP, but I think it tells us a little more about a start than SIERA or xFIP and I would have had to manually input all the kwERAs, which would have taken longer. Overall, I expected the consistency numbers to have a much better correlation than they did.

In conclusion, it doesn't really seem we can use any of these numbers to judge an individual start, at least from one pitcher to another. In future posts, I plan to look at some individual pitchers from start to start to see if these numbers have any correlation for individual pitchers.

No comments:

Post a Comment