The SABR Analytics Conference may have already wrapped up but that doesn’t mean there still aren't plenty of interesting and exciting things to cover. One topic that I found to be very interesting, albeit very complicated, is that of game theory and its application to baseball. Dave Studeman from The Hardball Times wrote a piece giving his thoughts on the SABR Conference, but more specifically there was a portion of that piece where he discussed the presentation of game theory by Matt Swartz.
At the SABR Analytic Conference, Matt picked right up with that future research thing. He presented data from 2-2 and counts and 3-2 counts and found that batters actually do swing more often at 3-2 pitches than 2-2 pitches (on pitches both in and out of the strike zone). He also expanded his analysis and found that pitchers and batters follow predicted behavior (throwing strikes and swinging at pitches) across all counts. Reality reflects theory.
Then Matt really dove into the data and found that baseball players aren't always maximizing their opportunities.
Most importantly, Matt found that fastballs are underused with no strikes on the batter and overused with two strikes on the batter. On the other hand, he found that batters swing too often with two strikes on them. Reality is no longer reflecting theory.
You can find all of Matt’s articles and his PowerPoint presentation linked to in Dave’s article and I strongly suggest you give them a read because there are some very interesting findings in his work. Up to this point, from what I’ve read about game theory and Matt’s work, I have a basic understanding of how it could be used and I’m confident it is the next big thing in how major league organizations view game strategy, but I also feel it’s still a few years or so away from being effectively leveraged because change doesn’t come easily or quickly in our sport.
I won’t pretend that I’m comfortable or knowledgeable enough to really hammer home the finer points of Matt’s research, especially the math portion, so I decided to reach out to Matt himself and get some questions answered.
Lance Rinker: This (game theory) seems like it would be a strategy if employed at the front office level on down, or even if some players took it upon themselves to try to really understand it, it could really up their game so-to-speak. But it seems like it is in its infancy when being applied to baseball, isn’t it?
Matt Swartz: A lot of people have tried it I think to smaller degrees but I don’t think they’ve done as full of a study both in terms of theory and taking it to data, and probably not as generally as I did. But I think that what I found is pretty actionable for teams or even a player that really wants to push it.
LR: Do you think that this is one of those things -- much like sabermetrics was early on -- where front offices are just now beginning to jump on board with advanced analytics and haven’t quite gotten to game theory yet?
MS: Yeah, I think that game theory has the potential to kind of be the next big thing, a place where you can take sabermetrics to strategy and make it really actionable but I think it’s a little bit harder than some of the other stuff that’s been done. Or not even harder so much as it’s just not in the wheelhouse of sabermetricians as much. I think that a lot of different sabermetricians have slightly different backgrounds than me, there’s not as many economists that do it so I think that it’s something hasn’t been looked into as much.
LR: Would you say that it’s specifically your background as an economist that allows you to apply this to baseball?
MS: I think it’s a combination of having the econ background and the sabermetrics background. There was a really good article that I mentioned in my slides by a couple of really good economists at Chicago, Kenneth Kovash and Steven Levitt, and they did some great data work and even more complicated data work than what I did.
But they used OPS and while that’s pretty good, the guys with .900 OPS’s are generally better than guys with .700 OPS’s, but it doesn’t value walks quite right so it didn’t really produce the kind of accurate results that it should have. So you need to know sabermetrics to know to use something like linear weights yet also know game theory but to think about it from a Bayesian perspective
LR: One of the things that stuck out to me the most, and correct me if I’m wrong, is that it seems like understanding this would benefit pitchers more than hitters. Would I be correct in that belief?
MS: I think so. That’s definitely the impression I gave off because that’s what I think. For batters there’s an element of they're going to swing if they think it’s in the strike zone and not if they don’t and they’re going to adjust their strategy a little differently earlier in the count, but there’s not as much of a discrete decision.
For pitchers, at any given time, they can choose to throw a fastball, curveball, or whatever other pitch they want to throw. So it’s a little more challenging for a hitter, I think, to decide whether they should swing if it’s this inch versus that inch off of the plate and this count versus that count. It’s a much more finite decision for pitchers.
LR: In your opinion, how much do you think psychology plays into game theory being effective for baseball since you’re dealing with people?
MS: I think that, ultimately, psychology is a big factor. It seems like you have hitters who are afraid of striking out looking. It’s kind of like a behavioral economics things where they distort what the best decision is based on errors in perception. And then from the pitchers perspective it seems like they don’t come back with the fastball in a 2-0 count as much as they should because they’re afraid of giving in.
LR: Another thing that caught my eye is the effectiveness of certain pitches. In your research, did you discover that maybe there are a set of pitches that are more advantageous for pitchers to learn and use in game situations over others?
MS: It’s kind of hard because I don’t know how easy it is for individual pitchers to learn individual things. I know that overall, the group of changeup pitchers in my study were much more effective than the guys who threw a lot of curveballs and a lot of sliders and sinker; it was a really effective group. I would guess that maybe there’s something about the changeup, but it could be it just happens to be a talented group of guys who learned the changeup because they had really good fastballs already and I don’t know if it would hold if you looked across multiple years.
There are only 10 guys in the sample with one year of data so it could move a lot. If I had to take a guess, with a gun to my head, I guess changeup pitchers but it’s really a guess.
LR: What’s the next step for you? Are you continuing to do more research with game theory and then apply it?
MS: Yeah, I’m going to adjust for some other factors and maybe look at some older data. I’d love the opportunity to discuss it with hitters. I think that getting some feedback from people about which type of pitchers to group together, whose curveballs are very similar.
I’d like to thank Matt for taking the time to speak with me about his work and continued research into game theory. You can find his work on FanGraphs, The Hardball Times, and his arbitration salary model is a staple on MLB Trade Rumors. You can also follow him on Twitter at @Matt_Swa.