The latest drop of the #TB12DB (or Tom Brady Data Biography) is live HERE. It covers the 2020 season and includes StatsBomb Event AND 30 frames per second Tracking Data. Unlike the last two releases though, there’s something a little bit different about the process used to generate this data than the earlier seasons.
This is the first public release we have done of integrated tracking and event data. Meaning all of the information collected here is collected out of the same process, not weaving together events and tracks by timestamp. In order to create this data set, we had to completely rebuild our entire data pipeline and collection tools to the new paradigm that locations = tracks.
Why should you care about it? Well, this process results in better quality than the competition and produces a ton of fascinating additional metrics you can create on top of the data.
And we’re turning all of this around in 12 hours after we receive video from our teams.
All of it.
Not some of it in 12-24 hours, and some of it on Wednesday… Our customers get ten times the event data, 30 frames per second tracking data, physical metrics, and the best quality in the industry… ready the next morning when they sit down to work. We have this data for the NFL and for FBS teams.
What you can do with event data and tracking in creating contextual physical metrics isn’t totally obvious, so let me tag in Faizan Subhani:
We are now able to generate speed, acceleration, and distance traveled for each player at the frame level (30/sec) against any all-22 film, that we have independently verified against NGS data (a full whitepaper will be ready for this in mid-December). In addition to providing the metrics that our competitors offer (top + avg speeds/accels), because our tracking data is integrated with our eventing data, we can contextualize these physical metrics in a way that isolates what matters on the field -- we can define the following football concepts more holistically:
- DB closing speed by only considering frames between when a pass is made to when the ball reaches the receiver
- DL get-off speed by measuring a defender's speed during the first second after the snap only on obvious pass plays
- WR go-route speed by isolating metrics calculations to only when receivers run routes where they're expected to go at top speed from the time of snap until the ball is thrown
Over the coming weeks, we're finalizing our initial approaches to measuring fatigue -- we'll start with receivers and the secondary and compare speeds/accels from the 1st vs 4th quarter in situations where we can expect them to be aiming to go as fast as possible. The next concepts after that are receiver speeds in and out of breaks and player load (as we will be able to combine our player height/weight metadata with frame-level speed/accel).
Want a visual representation of what DB Closing Speed looks like? Say no more.
Fine fine, but what does that mean for the #TB12DB Data Release?
Well, it means you can create your own hyper-contextual metrics from Tom Brady’s games off this new data in the same way we deliver for our customers. And you can compare it to the NGS Data from the NFL itself (and probably replicate any cool metrics they have been producing that piqued your interest).
As always, the data can be found on our GitHub page, where you will also find the 2021 and 2022 data releases, plus the R and Python guides for pulling and visualising the data.
If you do cool work with the data, let us know.
If you find problems with the data, let us know.
And if you want to start using the best tools and data in football for self scout, opponent scouting, and recruitment… definitely let us know.