Skip to content

SABR 40: New Technologies in Baseball Panel

New Technologies in Baseball Panel
Measuring ball flight using Sportvision’s PITCHf/x, HITf/x, and FIELDf/x
Trackman’s Doppler Radar Technology

Official description: Alan Nathan moderates a discussion of the latest developments in Sportsvision’s PITCHf/x, HITf/x and FIELDf/x, and TrackMan’s radar technology used to measure ball flight.

Dave Allen is an expert in spatial statistics and graphical analysis. He is a staff writer for Fangraphs and Baseball Analysts. He will talk about PITCHf/x analysis.
Josh Kalk is the Baseball Operations Analyst for the Tampa Bay Rays. Prior to that he was one of the leading PITCHf/x analysts and a writer for The Hardball Times.
Greg Moore is in charge of marketing for Sportvision’s baseball products, including PITCHf/x, HITf/x, and FIELDf/x.
Rob Ristagno is Director of Business Development for TrackMan. He seeks new markets in which to apply the TrackMan radar technology, including professional baseball.
Alan Nathan is an expert in the physics of baseball, with experience using both PITCHf/x and Trackman for trajectory analysis.

Alan Schwarz of the NY Times reported on a small conference on PITCHf/x and tons of people on SABR-L were excited about it and wanted to know if some of those topics could be covered here at SABR40. Alan Nathan was involved with the PITCHf/x conference and agreed to organize this panel.

One who was not listed in the program: Rand Pendleton is Director of Video Broadcast Development for Sportsvision and has been with all their projects. It appears he is replacing Greg Moore.

Rand goes first. There is an amusing and ironic pause as it takes the technology panel a small delay to get their Powerpoint running.

Sportvision started in 1998 and what they are known most for is that yellow line that shows up in football games. Then NASCAR that tracks cars within a few centimeters. Won some Emmys. 5000+ live events since 1998, not counting all the pitchf/x. Started out to enhance broadcasting, but a side effect was all the data collection, which is now a focus.

Pitchf/xis a camera based system, accurate to less than one inch.
Data is used for a variety of media analysis, teams, and sabermetricians.
All the pro broadcasts (ESPN, FOX, MLB, etc) are all using the pitchf/x data even if they call it something different.
Pitchf/x uses two cameras, 90 degrees apart, high home and high first. They are not genlocked. Installations at the MLB parks are permanent, so pitchf/x can have a broom closet to call their own. They run two computers, one for each camera, looking for the baseball constantly. They find one and then look for the next image and determine if it is a pitch. The software can tell the difference between a pitch and a hot dog wrapper. Then the images are carefully timestamped to give a trajectory series.
Can calculate Covington drag.
Speed, location, and trajectory are calculated immediately when pitch crosses home plate. Can replace the speed gun.

The cameras don’t see the ball at the same time, but they crunch the data. (Rand shows a nifty graphic showing the interleaving of the different image captures by the two cameras… I can’t really recreate it or describe it well.)
Keep the pixel density high enough for accuracy, but wide enough to see entire flight and 3D markers like third base.

Rand then shows some awesome graphs of pitchf/x analysis–which I can neither recreate nor adequately describe. You just had to be there.

Then comes HITf/x.
Uses the exact same cameras as pitchf/x, finds baseball in the images, and get initial speed, trajectory, and azimuth angle. Only tracks back to the pitchers mound so it doesn’t track where the ball comes down. (Unless it’s a bunt, obviously.)

Fieldf/x is a timed history of the players and the ball, two different cameras that show the whole field. Only one installation right now, at Giants PNC park. System will be able to calculate top speed of the player, range, and trajectories of the ball. Not completely automatic–needs some operator intervention but mostly automatic tracking of players and ball. In beta testing now so only operational in the one park, but two more will be up and running soon.

Question from the audience. When they show that pitch location graphic in two dimensions on ESPN or whatever, what is that meant to represent?

Answer: The box on EPSN’s pitch tracker, Gameday, etc… is meant to represent the front of the plate.

Rob is next to talk about Trackman
Incorporated in 2003 in a garage in Denmark, and has its roots in golf.
4 founders, 2 active in the company, started in Klaus’s garage.
In 2004 Trackman Pro demoed to customers, 5 demos resulted in 5 sales
By 2008 entered cricket, soccer, and baseball.

It’s military grade doppler radar software, so you get 24,000 samples in a single pitch. It locks on to the pitch as if it were a missile, and it follows it until it stops moving. Just one “camera” behind home plate.

The data it provides is actionable and the technology is scalable. “What happened?” (did the pitch break? and why?)

Actionability: release conditions, 3D release slot, speed at release, angle, spin rate, spin axis
plus the laws of physics, you will get movement (break horiz and vert), plate location, plate approach, time of pitch flight, speed at plate

So now you can diagnose if a pitch is different from another — did a guy change his angle? did his spin rate change?

Proves the more spin rate on the ball creates more strikeouts, so perhaps How able a guy is able to spin a pitch will be a tool that is graded in the future.

But spin isn’t enough if the spin axis isn’t good — one guy could get 2750 RPMs on his ball (MLB avg 2450) but if spin axis not good ball won’t break or have bite.

Also the lower the vertical release angle on a curve ball the more swings and misses because of deception. You have to release it a higher angle than other pitches, but you can tip your pitch if you have it too high.

Likewise, Trackman can tell on hits the launch angle etc.

Well hit versus not well hit shows hot and cold zones on data compiled into chart. Another one of those graphics you had to be in the room to appreciate.

If a batter is in a slump, you might be able to actually tell if he’s not getting the right exit speed or launch angle or if he’s just unlucky.

Right now the data is proprietary to Trackman customers. But Josh Orenstein runs the Trackman Baseball Insight Lab, jko@trackman.dk and you can get your hands on some data through there and your ideas.

DAVE ALLEN
Using pitchf/x to measure pitch success by location
Can use pitchf/x data to assess pitch success by location where it crosses plate (or where it WOULD have if not hit before getting there)

Assign linear weights run value to each pitch: what was the change in the expect number of runs before and after the pitch
Scale pitch height within the strike zone to account for different strike zone sizes

Fit a loess regression surface to these values

For more info on this method and replicating the analysis, see The Hardball Times Annual 2009, Dave’s pitchf/x Summit 2009 talk, and his posts at Baseball Analysts

A strike it about -0.05, while a ball is about +0.05

Shows some cool diagrams showing things like Dustin Pedroia’s hot and cold hitting zones. Again, you had to be there.

Since pitchers do not have perfect location this isn’t as actionable as you might always like.

One thing talked about is tracking the catcher’s glove and how good are guys at hitting their location.

You do have to look at one standard error from Pedroia’s graph, since a single year’s sample size for one batter. is small.

You can get all the pitchf/x data for free from MLB Advanced Media in xml, of through various third-party online providers (brooksbaseball.net, texasleaguers.com, joeleftkowitz.com)

Even some players like Max Scherzer and Brian Bannister are looking at their own pitchf/x data.

Josh Kalk “The Red Dot”
(Amusing, Josh has put the date 9/7/2010 on his talk. However it is only 8/7 today. Numbers geeks, I tellya…)

Begin by playing a Reggie Jackson quote from NPR.
“If you can’t see the rotation, you have to be able to recognize if it’s a curve ball or a slider. And if you can’t, you’re not going to be a major league player. Anyone who can hit above .270 can see the red dot. Anyone who saw that dot on the ball, you knew it was the slider. If it was a really big dot, you knew it was a hanger and you could hit it out.”

Reggie claims that small dot means good slider, large dot means a sloppy one.

Josh then shows, using a baseball attached to a dowel, spinning, as a four seam fastball. Pure backspin. It gives a rising action (counteracted by gravity). Then he puts them on the graph he’s making.

Then he shows a curve ball, and places them at the bottom of the graph. Neither of these has much left/right movement.

Then he shows a ball spinning as if its drilling toward us with a gyroscopic movement — like a football or a bullet. It gives no effect left, right or up or down.

Finally the slider has a “drilling” motion toward the ground which gives it left/right motion.

The dot comes from the red seam always ending up on the spin axis.

Josh then produces a power drill with a baseball attached to the bit. So he can show everyone the red dot. This is fabulously successful until the ball flies off the end. Um.

“Apparently a little more epoxy was needed,” he says. But the point is made.

(I should point out that the red dot and “nickel” curves and so on was covered quite a lot of this topic in the Baseball Research Journal, oh, now Alan Nathan is mentioning that Dave Baldwin and he worked on the article “Nickel and Dime Curves.”)

ALAN NATHAN, four quick things
showing nifty graphs using these kinds of data to answer questions

Why Is Mariano Rivera So Good?
Location, Location, Location.
Shows a graph built from pitchf/x data that shows that he “lives on the black.”

“Late break” truth or myth? Graph shows a 5″ break from the straight line trajectory, another shows less that 1 inch of break. How is the poor batter to know which one he’s getting? Shown on actual trajectories. It just appears to be a late break when he gets the 5″ one.

Using HITF/x on BABIP establishing outcome-independent metrics
HR have a launch angle around 30 degrees. But if you want to get on base, you want to have a smaller vertical launch angle. Angle correlates with outcome. High batted ball speed and vertical launch angle 10-12 degrees gives a line drive.

Combining HITf/x with Hittracker
HITf/x gives initial sped and direction
HIttracker (Greg Rybarczyk) records landing point and flight time of every batted ball
Together these constrain the full trajectory!

So, does the ball carry especially well in the new Yankee Stadium in 2009? were there funny wind currents?
“carry” = actual distance/vacuum distance
so a ball that went 397 feet but would have gone 571 in vacuum, has a carry of 0.695
Found the average “carry” at all ballparks, so normalized to one. Those with higher than average carry would be greater than one.

Denver has a carry of 1.07 — the highest of all. Texas also high, 1.04.
Cleveland has the lowest, .98. San Francisco also pretty low.
Yankee Stadium came out slightly below one, as well. Totally average. Also didn’t find an effect looking just at right field of Yankee Stadium either.

(In other words, it was the ghosts of Ruth and Gehrig coming over to the new stadium making it seem like that and now they’ve settled down. Wink.)

Questions from the audience.

Post a Comment

Your email is never published nor shared. Required fields are marked *
*
*