MLBAM Player-Tracking System Is Set for Coming-Out Party This Year
Last year, MLB Advanced Media stirred up interest among fans and broadcasters alike when it launched a data-driven player-tracking system at three ballparks, aimed at measuring the “game of inches” in a whole new way. When Opening Day arrives in April, that system will be up and running at all 30 MLB ballparks, and fans will see the data-visualization graphics on national and regional MLB telecasts as well as on MLB.com and the MLB At Bat app.
“It all starts with making sure you have the most accurate data possible,” MLBAM EVP/CTO Joe Inzerillo said during a case study at SVG’s Sports Graphics Forum on Wednesday. “Context is everything, and, if the data is not consistent and accurate, you have no context, because you can’t say this guy was better than that guy.”
Capturing the Data: Integrity Is Key
Relying on ChyronHego to handle optical tracking for players and throws and Trackman for radar tracking of pitch, hit, and throws, MLBAM is responsible for calculating all metrics based on the raw data (vendors are not involved in developing metrics).
“We are using both [ChyronHego and Trackman] systems to hint each other in the throws,” said MLBAM VP, Multimedia Development, Dirk Van Dall. “Radar is very good at arcs and high balls, but it essentially loses the ball when it hits the ground. Optical tracking can hint at where that position is, so, working together, we can understand where the ball is and verify the quality of both data sets.”
The system — which debuted last year at Miller Park in Milwaukee, Target Field in Minnesota, and Citi Field in New York — also takes into account error mapping, confidence scoring, and the raw-data provenance to ensure that the data is as accurate as possible.
“We wanted to make sure we captured the confidence we have when sampling various data and parts of the field,” said Van Dall. “Some of the data comes from optical capture, other data from radar capture, and some is extrapolated when we are missing samples. We are making sure that we understand what the quality of that data is now so, in the future, as our systems improve, we can come back and look at the data for this year and refactor it.”
All data points from each play are segmented into a single data set, aligned with the time stamp for the beginning and end of the play, with the standard-English description of the play.
“This [data] can then be displayed as a replay [immediately] or looked at intra-game or season-wide,” said Van Dall. “That is really where you are going to get the storytelling: understanding why a certain play is exemplary, because you can compare it to other identical plays or matchups over the course of the season.”
Using the Data in Broadcasts: Real Time vs. Instant Replay
A wealth of player-tracking data and graphic elements will be available to the MLB national and RSN broadcast partners this season. MLBAM breaks the metrics provided to broadcasters into two categories: those available in real time and those available with a 12-second delay (similar to the time frame of an instant replay). Inzerillo says that many of the features seen in Statcast VOD segments released last season will now be available to the live production.
“A lot of what people have seen us put out as VOD clips after the event were done using the same mathematical rigor that we are now going to be doing in real time or within 12 seconds,” said Inzerillo. “What you are going to see [this year] are things that we can actually render in 12 seconds. This is a combined precomposited playback, so it’s essentially like driving an EVS [replay server] with data overlays popping in automatically.”
Breaking It Down: Pitching, Batting, Base-Running, and Fielding
In terms of pitching metrics, actual velocity, extension, perceived velocity, and spin rate will be displayed in real time, and metrics like holding runners and pickoff timing will be available following a 12-second delay.
For batting, real-time metrics will be exit velocity, launch angle, projected home run with distance, hang time, fly-ball distance, and vector.
During last year’s MLB Home Run Derby at Target Field, MLBAM’s prototype system accurately predicted all but one home run out of 200 total bombs (the single outlier hit the line on the outfield wall) while the ball was still traveling over the infield.
“One of the other cool things is flat-carry distance,” said Inzerillo. “It’s not just how far the ball was hit; it’s how far the ball would have gotten if there wasn’t a structure in the way. Most of the time in baseball, we have been calculating HR distance based on where the ball landed in the stands, but that doesn’t give you a real approximation of power — especially when you start to go ballpark to ballpark since the walls are in different places.”
Real-time base-running metrics are lead distance, acceleration, max speed, and home-run trot. First step, route efficiency, stealing first step, and secondary lead distance will be available with a 12-second delay.
“Base-running metrics is where you start to get into things that are really cool,” said Inzerillo. “Unlike pitching and batting stuff that builds on things that have come in the past, some of the base-running data was simply not captured in any objectively quantifiable way in the past.”
Fielding metrics available live will be acceleration, max speed, and shift positioning; 12-second–delay metrics will include total distance on caught balls, first step, arm strength (catcher and fielders), exchange (catcher and fielders), pivot, and catcher pop time.
“Part of the [fascination] of professional sports is [in] making the impossible look easy and the easy look impossible; now you can actually figure out which of the two a player is,” said Inzerillo. “There are a lot of players that get terrible jumps but make spectacular plays because their athleticism can make up for the fact that they didn’t read the ball well. Likewise, there are other guys who make it look rudimentary, but, when you look at the hardcore metrics, you see this guy can really cover some distance. That ability to quantify over the course of time to build context is really one of the most important things that we are trying to do here — especially on the defensive side of baseball, which has never had great statistics.”
Looking Ahead: Automated Cameras, Device-Specific Content
While much of the focus this season will be on integrating the player-tracking elements into the broadcast workflow, Inzerillo sees this as one small piece of both the production and the consumer sides of the business. He recounted an experiment that MLBAM conducted last season in which player-tracking data drove high-frame-rate pan-tilt-zoom cameras for close plays at the bases. The cameras read the players’ location and moved proactively to the base where there was going to be a play.
“That is a pretty good example of the type of integration that is also possible,” he said. “It may not be purely statistical; it could also be contextual for pan-and-scan cutting or for driving physical devices on the field because you don’t have to guess when you operate; you can just automate it. It’s coming, but how fast depends on the traction and some other factors.”
On the consumer side, Inzerillo sees the potential to use this mountain of data to drive consumer experiences beyond the linear telecast.
One example could be wearables. Inzerillo foresees a scenario in which MLB’s iBeacons at a ballpark can detect a fan’s wearing a smartwatch. If a pitcher throws a wicked fastball to end the inning, the fan could potentially use the watch’s glance action to immediately see the speed and path of that pitch.
“That is a great example of how it’s less about putting out a big data set than putting out the right data set depending on the context of what the customer is consuming,” said Inzerillo. “This is what we are going launch the season with and you’re going to see in broadcasts, but this is 1% of where we think this is going to go over the next couple of years.”