clock menu more-arrow no yes

Filed under:

Zebras, Technology, and Next Gen Stats in the NFL-Part 2

New, comments

Heading into the bye week, it’s worth looking at how Next Gen Stats are viewing the 2018 Bears.

NFL: Tampa Bay Buccaneers at Chicago Bears Kamil Krzaczynski-USA TODAY Sports

Zebra Technology, Part 2

On Wednesday I shared my impressions of the people and the technology behind the NFL’s Next Gen Stats, Zebra Technologies. With the basics out of the way, and with the 3-1 Bears headed into a bye, it’s a good time to look at what that technology is telling us about the Beloved. Along the way, though, there are also a couple of interesting points to make about how this technology can (and cannot) alter the fan experience.

Draft Day Anticipation

One time-honored tradition in the NFL is the Mock Draft. In fact, the NFL Draft has become almost its own performance and game, similar to a sport of its own category. GMs compete more directly, only in a manner that is harder to evaluate than with a simple scoreboard.

One obvious implication of Zebra Technologies’ tracking data is that they can give a much richer data stream to the GMs are who trying to find the best possible players. It’s already fairly well-established to those who look into the data that the NFL Combine itself is a flawed tool of evaluation. However, one of the simplest rejoinders is “what else is there”?

Zebra Technologies might have the answer, because they’ve already been invited to the Senior Bowl to track players and practices, and they are working on expanding into other bowl games when they can.

This means that if a GM could figure out which piece of data is meaningful, that GM could capture a brief advantage for a couple of years until other front offices caught on. In the NFL, a two- or three-year lead on finding the right players could mean a difference between landing a few starters or finding the next third-round All Pro.

“Charting” Trubisky’s Development

One of the nicest things about having clean tracking information is that it lets fans evaluate their own impressions. In 2017, Trubisky’s “deep” ball--passes over 20 yards--were all pretty solid (his deep ball to the left actually had a 118.8 passer rating). A simple look at this chart might cause someone to think “well, he could always throw deep and to his left just fine--false narrative!

However, his “intermediate” passes to the left (passes to the line of scrimmage and up to about 20 yards) were all well below the league average. More than that, actually digging through the charts (not the summary, but the per-game charts themselves) reveals that he did have a problem. He didn’t throw deep and to the left often, and when he did the balls were incomplete a fair amount. However, he did have a single touchdown strike in that range, and that tilts the numbers. A specific look at the 2018 charting tells us the exact same thing--the didn’t make such passes often, and suffered incompletions on many of the efforts when he did try, but then he had a handful of truly impressive throws against the Bucs.

None of this is intended as an indictment against Trubisky--he improved his performance and time will tell if he can sustain that level of improvement. However, it does reveal something important about Next Gen’s advantages and disadvantages. It is absolutely a richer representative of a game than a box score. However, just like a box score, its summaries need context to be meaningful. It takes looking at the details behind the numbers to really be able to use them. For example, clicking on Tom Brady’s chart reveals he doesn’t even have passes deep and to his left yet, this year.

In the case of the Bears, Trubisky is developing, but the “narrative” about his struggles has some evidence behind it, too. The curious fan just needs to look at the actual game charts instead of the snapshot summary.

Holes in the Coverage?

A Bears fan looking for bragging rights might want to go to one of the other sections of Next Gen and find yet another superlative for #52, Khalil Mack. Surely this Monster of the Midway has one of the fastest sacks, right? Nope. In fact, no Chicago Bear has recorded one of the fastest twenty sacks in the league this year. That seems almost strange, given that Chicago leads the league in total sacks. Except, really, the Bears are getting their sacks with discipline and persistence. The whole Front Seven is working together and Mack and Co. are being relentless in their pursuit. In other words, they aren’t accumulating sacks as a result of a handful of freakishly athletic plays. They are, instead, dominating this category with consistent skill and solid play on all levels. Sacks are coming from defensive backs stymying the quarterback and from pass-rushers working together as a team.

This sort of thing still doesn’t show up on Next Gen’s public tables, even though it’s the kind of data they probably have saved somewhere (how many players were within a certain distance of the quarterback at a given moment, or how open particular receivers were when a sack occured). In other words, while the action of a game can be reconstructed using this tracking data, it is simply another way of getting information, and that information is so rich that it has to be moderated somehow. Some of those limits (like tracking fastest sacks but not having a metric for ‘coverage’ sacks, for example) are why true Xs and Os analysis is going to stay a part of football coverage for a long time.

Here’s another example--not one of the 20 fastest “ball carrier” plays (meaning the peak speed reached by a player while carrying the ball) belongs to Tarik Cohen. Tyreek Hill’s peek 21.95mph touchdown reception in Week 1 carries top honors, and the fact that Jordy Nelson (now of the Raiders) has two entries on the list suggests he hasn’t lost a step at all. However, Mr. Mismatch isn’t on the list. Why not? My guess is because he is more shifty than explosive, and more agile than fast.

This means that the data Zebra Technologies is providing makes for good talking points, and it can provide excellent clarification on a number of levels, but it is still only as good as the questions we ask it.