The State of F/X: MLB.com Gameday and Baseball Analysis Improves

Harry PavlidisSenior Analyst IMay 12, 2008

The work many folks have been doing with PITCHf/x has been great. It's getting better.

This weekend was the 1st Annual PITCHf/x Summit. We all hope we'll have a 2nd one, too. So many topics were discussed, it will be difficult to put it all together, but I'm going to try.

The event kicked-off Saturday May 11 at the Westin on Market in San Francisco, just a half mile from AT&T Park. I arrived Friday and had the pleasure of attending the game with Greg Rybarczyk of Hit Tracker. While folks like me analyze, Greg innovates. We were both bemused by the amount of drunk late-teen/early-twenties in the left-field bleachers. No, we're not grumpy old men, this was like nothing I've quite seen before, in terms of early drunkenness and fighting. I've been to the bleachers in Wrigley, the Blue Seats in Madison Square Garden, and many points between. Friday night was certainly new for me. As a matter of fact, the night ended tragically, when a teenager was killed after striking his head on the concrete, following an alleged punch from behind. The deadly blow was allegedly dealt by another teenager. Both had graduated from local high schools in the past couple of years. This is the only black mark on what otherwise was a great weekend, which I enjoyed beyond my own expectations. 

Blissfully ignorant of Friday's tragedy (Greg left a comment on Cubs f/x, telling me what happened, Sunday night), we had a great time at the conference, and all overcame our guilt of being at a baseball conference on Mother's Day. Mothers with broadband connections could tune-in on-line, however.

Much of the discussion, as you can imagine, was highly technical. Math, aerodynamics, high energy physics software applications, databases, scripting languages, user interface, human factors, neural networks, data correction algorithms, lens distortion, camera characteristics, and my talk on creative uses of the data. Decidedly non-technical, I hope I provided an informative and entertaining look at the world of PFX. But, probably not. You can see for yourself, as the presentations are available on-line, and, eventually, a down-loadable DVD might be available, too.

Matt Lentzner had a baseball handy throughout, and some thought provoking innovations on arm slot and pitch identification. This is a topic of significant interest to me. At the ballgame Saturday, I got nicely sunburned while Matt, Alan Nathan and I discussed the matter, and Matt and Alan both used the baseball to explain and explore ideas. Wisely, I stayed on the bench, hand on beverage. Matt presented, at the last minute, his related ideas on pitch break distribution as a tool to find arm slots, and, in a few quick slides, showed his suggestions on how to show pitch movement in Gameday to make it more intuitive to the casual fan. The MLBAM guys were very interested, and a very good discussion followed.

MLB Advanced Media provides us with the great stuff in Gameday, with the new 3D version coming very soon. We did get to see it in action on Sunday, and they've done very nice work. I think you'll all be pretty happy with it, too.  MLBAM was represented by Ross Paul, who provides us with those real time pitch ID's, and Justin Schafer. Justin and Ross shared a lot of information, context and perspective on the whole endeavor. And they listened. They asked questions, answered questions, and were fully engaged in the conference. Along with a gaggle of great people from Sportvision, a few spare physicists, and a bunch of bloggers, were representatives from nine major league teams. Including the creator of favorite baseball simulator, so I got to have my geeky fan moment.

Going back to Greg and his outstanding site Hit Tracker, I want to talk about future possibilities for the system. In particular, Hit f/x. With the existing cameras, you can pick-up batted balls - from high-launch homers to Baltimore chops - and Peter Jensen as the pictures to prove it. Peter got a collection of video files from Marv White (CTO, Sportvision) and, along with Rand Pendleton from Sportvision, was able to put together a variety of launch metrics from the existing tools. Alan Nathan, of the University of Illinois, had already done a study combining "hit f/x" data with Greg's data, and it seems they're on to something.

Besides having good company, a benefit of attending Friday's game with Greg was learning more about his work. I also was able to witness his process of photo-documenting nooks, crannies and landmarks. Greg's model of AT&T park is about to be improved, just in time for some Barry Bonds analysis that Greg may be about to embark on. Greg's talk on Saturday integrated fielding data via examples taken from a webcam in Nationals' Park. No, the cam is no longer operating, but he was able to find at least four shots that captured a ball in flight (one appears to be catcher's toss back to the pitcher) and lots of information on fielder positions. The work on home runs that we're all so familiar with is going to expand to more and more batted balls. Greg sees no technical reasons to not do this, only practical limitations and considerations raise doubt about the feasibility of the "Ultimate Gamecast". Greg not only gets this stuff technically, but he fully grasps the "immersion" factor that is so crucial to all fans, casual, technical, and, I'll throw in newbies, too. Be sure to visit the Summit web page and get Greg's slides. Like I said, innovations.

Now, time for the foam board. Sportvision's high tech lab (parking lot) and several baseball diamonds have played host to their engineers on many occasions to test the system. Setting up the PFX cameras, and placing a foam board on the front-plane of home-plate, they have been able to establish the level of accuracy of the px/pz values. Within 1 inch. Sometimes better. In any case, you and I can trust those plate locations within a third of a baseball. That's not enough for me to care about, so you'll see me exhibit great confidence on "red meat", "corner nibblers" and "ankle burgers" diagnosis going forward. Almost the same with the break numbers, too.

While the accuracy of the plate location is so precise, it makes you shake your head, the pitch tracking itself, from release all the way to plate, is nearly as good. Alan Nathan has calculated an approximate 2.2 inch noise factor in pfx_x and pfx_z (not the same for each, I believe). He has an improved approach that tightens that up a bit, and I believe Ike Hall's ideas on data normalization along with a lengthy discussion of the coefficient of drag will take Sportvision closer to even better data quality. But, let me tell you this much, this stuff is tight. I can't say "this curveball broke down 6 inches" for two reasons - it could be four or eight, and, Matt Lentzner would want me to add gravity and then adjust back to the hitter's reference point. But I can say, with the exception of Matt's idea, which I think will become the way of the future, or close to it, "his average curveball breaks down 6 inches". I got to see the operations with my own eyes, and they super-impose actual and calculated trajectories on the display, and, wow, this stuff is pretty tight.

There are, however, issues. Camera positions vary, since they need to be placed such that the screen/net is not in the way. They don't pick up all the way to the rubber (or, at the ground at least) to conserve pixels and get the richest data possible on the flight. These are amongst potential factors leading to variance in release points. I also include mound height in there. I also had the chance to point out the Great American Ballpark anomaly, about 7 inches of distortion in spin movement, to Sportvision, and realized (a) my own obligation to provide the feedback and (b) their thirst for it. Good scientists work transparently, and it produces better work.

This brings me to Kerry Wood's back-up slider. I have to take in some more data, but, so far, I do believe I can see it (I did show it at the Summit, so you can download the slide), but the GAB effect and the small sample don't allow me to declare it so. But, I think I'll find it clearly enough, soon enough, to make that declaration.

Marv White did point out we are asking the piano playing cat to sing, but he seems to be driven by innovation as much as anyone. He is a business man, too, and that's a key to this. It is a business venture, which I was pleased to learn more about, and developed an even greater understanding and respect for the work done by all the parties involved, from the stadium grounds crew, to teams and the league. So, I'll end this by thanking Marv, Justin, Ross, Rand, Catherine, who deserves to be thanked twice, so I will, Catherine, Greg, Peter, the Lookout Landing guys, Mike Fast, Max "Professor Pepper" (come back to the US and sign that ball, please), Ike, Tom, Zach, Helen, Alan, Mont, along with a great high school physics teacher from San Mateo, and everyone involved in the Gameday, the Summit and PITCHf/x. Apologies to everyone I forgot.