Monday, May 16, 2016

Searching for the Next Frontier in Pitching Analysis

Baseball statistics have come a long way. We've reached a point where things like OPS and FIP have become almost universally accepted, and where the analytically inclined can easily access stats like wRC+, SIERA and WAR.

We've also come to a point in time where baseball data is at it's zenith. With things like Pitch F/X and Statcast giving us the most granular and comprehensive data we've ever had, there's seemingly nothing we can't track.

The problem is, we don't quite know how to use it yet. Baseball is such a complex game that even with all the data we could imagine, we can only scratch the surface of what it means.

I want to write briefly about a few ideas I've had that could possibly help us use Pitch F/X to deepen our understanding of the art of pitching, and ultimately improve our ability to analyze it. I haven't researched any of these ideas, but I'd like to at some point in the future.

Idea 1: Identifying pitches that moved in to or out of the zone

In general, if a pitcher throws a pitch in the zone and the batter takes, it results in a good outcome for the pitcher (a called strike). If a pitcher throws a pitch outside the zone and the batter takes, it results in a bad outcome for the pitcher (a ball). Generally speaking, pitchers want batters to chase pitches outside the zone, and to take pitches inside the zone. When I wrote about Joe Ross, I wrote about a metric I called Fool%, which attempted to quanify a pitcher's ability in this area.

Now, hitters are much more likely to chase a pitch out of the zone if it looks like it's headed for the zone and dives out of it at the last second (think of a batter fishing for a curveball in the dirt), and are much more likely to take a called strike if it appears to be outside the zone before darting into it at the last second (think of a pitcher freezing a batter with a frontdoor fastball). With Pitch F/X, we have the necessary location and movement data to identify pitches that looked like strikes but weren't, and pitches that didn't look like strikes but were. My guess is that the percentage of such pitches a pitcher throws will correlate very well with K%, as well as Fool%. The percentage of pitches within the zone that looked like balls should have a significant negative correlation with Z-Swing%, and the percentage of balls that appeared to be strikes should have a significant positive correlation with O-Swing%.

The biggest hurdle in testing this theory is determining what the vertical movement reading would be for a hypothetical straight pitch, since the movement readings in Pitch F/X are compared not to a straight pitch, but to a pitch with no spin.

Idea 2: Difference in movement between pitches (also known as "tunneling")

The thinking in this one is that, if a pitcher throws two pitches that move in opposite directions, it's impossible for the batter to cover them both. The idea is to look at the average movement and velocity of a pitcher's pitches and determine, if they were thrown at the same trajectory, how far apart they would end up. I'm not entirely sure what this would correlate to (perhaps Contact%? Or maybe soft contact? Or maybe both), but tunneling is certainly an important part of pitching; when the batter doesn't know what direction the pitch will dart off in, it's an additional advantage for the pitcher. Additionally, it shouldn't be too hard to come up with, and wouldn't require mining through the expanses of individual pitch data, as you can simply look at average movement by pitch type. It could also help a pitcher expand the zone even more. If the batter thinks the pitch is going to move into the strike zone he may swing even if it was outside the zone the whole time and end up chasing a pitch that moves even farther from the zone.

Idea 3: Average difference from pitch to pitch in effective velocity within an at bat

The idea of changing speeds has been around forever, and August Fagerstrom of FanGraphs did a solid job quantifying it in an article on then-free agent Wei-Yin Chen. But the idea in my head at that time that has stuck with me is that instead of using raw velocity, the data would be more instructive using the concepts of Effective Velocity

If you're not familiar with Effective Velocity, I highly recommend the linked article, which was a great read, but I'll try to sum it up here:

To hit a pitch up and in with authority, you need to be a little bit out in front of it, or you'll just pop it up. Since you have to start your swing earlier to make solid contact, the pitch's functional velocity (or "effective velocity") is faster than it's actual velocity.

To drive a pitch that is on the outside corner and at the knees, you have to be a little behind it or you'll roll over on it and pull a weak ground ball, so it's effective velocity is slower than it's actual velocity.

By combining this effect with actual changes in velocity, you can amplify the effect of changing speeds. I'm betting a stat like "average change in effective velocity from pitch to pitch within an at bat" has some significance, whether it be in generating soft contact or in getting whiffs (or both!)

No comments:

Post a Comment