Every year the NFL runs a contest for stat nerds everywhere, known as the Big Data Bowl. They provide the data, the rules, the question, and $100,000 in prize money. Last year’s big question was about predicting the result of a running play based on information available at the time of the hand-off. This year’s question was to discover what makes a successful defense during passing plays.
Winners were announced on Friday morning (the contest is judged by NFL personnel) and one in particular caught my eye. The winning entry by the group of Asmae Toumi, Marschall Furman, Sydney Robinson, Tony ElHabr, is elegant, easy to understand, and they produced an easy-to-use, searchable database of their findings.
How it started How it's going pic.twitter.com/VRHzSvg9ZC— Computer Cowboy (@benbbaldwin) February 5, 2021
Toumi’s team named their metric after Wade Phillips, which, if nothing else, deserves points for proper marketing in a world of DVOAs and EPAs. WADE (Weighted Assessment of Defender Effectiveness) focuses on the secondary, and breaks pass defense into the distinct categories of “coverage” and “contest.” The coverage half seeks to identify those pass defenders who excel at not being targeted in the first place via playing tight coverage. These are your Deion Sanders types.
The second metric attempts to determine how good or bad the pass defender is at breaking up throws when they are targeted. Coverage is measured by a metric called dTPOE, or Target Probabilty Over Expectation, while the ability to contest catches is measured by dCPOE, or Catch Probability Over Expectation. (The name here may need some work as CPOE, or Completion Percentage over Expected, is already a well-known stat.) The gory math as to how they arrive at these metrics is all available here and I’d urge you to take a look, as this group is as talented at explaining their work as in doing the work.
They also helpfully provided a full database of how every pass defender ranked in 2018, meaning we can look at the Packers secondary from that year. For context, the highest-ranking corner by dTPOE was Kyle Fuller of the Bears at -20.34 percent. The best player contesting throws was Justin Coleman of the Seahawks at -5.385 percent. The two best secondary players overall are Fuller and Stephon Gilmore of the Patriots. You can see a breakdown of safeties and corners here.
The Packers are an interesting crew. The 2018 Packer pass defense wasn’t good, but they had just recently drafted most of their current corners, and they were on the verge of making a big switch at safety.
2018 was Jaire Alexander’s rookie year. My lasting memory of that year was of Alexander always being tight in coverage, but also getting beaten for a good number of big throws and deep balls where he was in good position but lost at the catch point. This ends up being pretty dead-on, as Alexander had a very good coverage dTPOE of -11.02 but was second-worst in the league (222nd of 223) in dCPOE at 5.137. Alexander has some good company with that split, as Marshon Lattimore, then in his second season, is almost identical, as is Josh Norman. While we only have 2018 data to work with, I think it’s safe to say that Alexander has greatly improved this facet of his game while not losing anything (and likely improving) in the “contest” area.
As for the rest of the Packers secondary: 2018 was also the rookie year for Josh Jackson, and it’s worth noting that Jackson may have been better than Alexander that season. His coverage grade of -9.21% was 58th, only marginally worse than Alexander at 45th, but Jackson was much better when contesting passes, ranking 110th with a dCPOE of +0.044. Jackson has had his issues since his rookie year, especially with penalties, but I thought he played well in the middle of the 2020 season before finding himself in the doghouse for unknown reasons. It’s worth remembering that Jackson is talented, and has shown it at times.
The Packers also briefly employed current Kansas City Chiefs cornerback Bashaud Breeland in 2018, and it’s interesting just how similar he was to the 2018 version of Kevin King. Breeland had a -4.26 dTPOE (coverage) grade, ranking 104th, while King’s was -2.74, ranking 122nd, but the real similarity was in contesting balls. There, Breeland ranked 32nd while Kind ranked 31st with essentially identical scores. Tramon Williams, in his aged form, was no longer great in coverage ranking 168th, but he still excelled at the catch point, ranking 40th.
Everyone else was varying degrees of bad, which makes perfect sense as the Packers’ defense overall was atrocious, though you can see some of the logic to how they might one day grow into something more. If Alexander was to be the alpha, excelling in taking away targets as he does today, it makes sense to employ other corners who excel when a pass does come their way. You can’t have corners that are too bad in coverage, of course, but it does make some sense to have them excel when they are targeted.
I’d wager that King has fallen off a cliff across the board in the two years since this dataset, but his lofty contested catch metric does make some sense, as he has ranked as one of the best corners in man coverage in the past and has been especially good in the red zone as recently as recently as last season. I suspect injuries have taken their toll on King, and it’s worth remembering he was questionable entering the NFC Championship game and likely played injured.
Finally, it’s worth considering the big question at safety. In 2018, Adrian Amos was still with the Bears, and while Chicago’s best and brightest assured us that Eddie Jackson was the real star, in reality, Amos and Jackson were nearly identical in coverage. Jackson was 75th in coverage, while Amos was 80th. Jackson was 4th in contesting catches, Amos was 8th. The Packers chose wisely, especially when you consider that Ha Ha Clinton-Dix ranked 88th and 115th, respectively.
This was hardly the only interesting Big Data Bowl submission. There is another that ranks pass defenders by their susceptibility to single vs. double moves that is also quite interesting, and I’ve only scratched the surface of my reading. If you have any curiosity about the state of analytics in sports I’d urge you to check out the winners. It’s actually incredible that the NFL does this at all, and it’s fantastic that the results are public. I hope this particular metric (or something like it) manages to find its way to the Next Gen Stats page.