Eric Eager joins the RAS podcast to talk NFL Divisional Weekend
This week's guest on the RAS podcast was Eric Eager, VP of Research and Development for SumerSports. Eric leads a team of data scientists who provide data and strategy to NFL teams.
On this episode, he was asked a series of questions (partial transcript below) about what it is like to work with NFL teams, his experience with player grading systems, roster/positional value, in-game strategy and of course, all four games this weekend.
Listen to the full episode here:
Are teams coming to you guys, what is the relationship like on the business to business end?
SumerSports is a little bit different than Pro Football Focus. At PFF, we were a data company first and an analytics company second. We were tasked with selling a $250,000 product to analytic staffers in the league, it was a hard sell, but a great triumph to get all 32 teams on board. At Sumer we are often approaching teams ourselves though. We have a few pilot clients and it is going well. But we think it will go both ways. Ownership will push teams into this, and we will have to approach with unique sales, trying to address what their needs are.
Can you hint which teams you work for?
No, I can’t [laughter], but it is funny, it was the same thing at PFF. But you look at other competitors like StatsBomb and Ted Knutson would always be tweeting out when they got a new client and I think it is awesome. In football though, I do not know why it is different. Teams are cagey about this type of thing.
Can you tell us which teams for certain you have not helped?
Uh….I can’t either [laughter]. It’s funny, because there are some of the smarter teams - some of them play this weekend - one would think those are certainly going to be Zelus or Sumer clients, but a lot of teams like to do things in house and keep expertise there. But then you have other teams who go with an outside vendor, and they will go all in thinking if they are using a product everyone else is using it will help them keep up. Analytics are still a second order term to getting a head coach and quarterback, if that makes sense?
If every team uses your product, are you not devaluing the product by educating the league with the same stats and info?
I think so, yes. But one of the big differences is that we will let teams give us their own grades. Teams loved data, they loved [PFF] Ultimate. They would log in, look at every play when Marquez Valdez Scantling ran a slot route, and you could see all of it. What they did not love is the grades. They did not love someone on the outside how well their players played. Now, some teams did love it, college teams - some of them - used their grades as their only scouting system. So for us now [at Sumer] to come in an optimize their roster around the teams own grades, it will create different optimal. If you are a team like the Ravens you value things differently than the Chiefs, so you will want different things. With football there is such a local element around strategies, but yeah, there is always a worry about your question. A company like Zelus for example, only sells to one team per division. It’s an interesting model and one to talk about.
Let me ask you about the grades, if you are doing 130 college teams, 32 NFL teams on any week, you’re trying to look and grade 10-12-15,000 plays. How is it possible to keep up? How can there be that much trust?
Well, there is a rubric. So I would say, for a Sunday Night Football game, for example, there are 15 data collectors who collect it live and get it to Chris in the booth. So as someone who helped with that it is a big undertaking. For a college game, it is lower, about 8-10. About 2 people to do base data, how long was the drop back, how long did it take the punt to get off. Those are double blind, so they are reviewed. Same thing for the graders and participation. Those are all coming from a rubric. So that gets you to 500+ data collectors, those are long weekends for those people, for better or worse.
I think personally there are holes in the system. We saw Patrick Mahomes get the 8th highest grade in the league one week when he threw 5 touchdowns. I think as the game evolves, you have to evolve the grading strategy. You can chart different data, like if you’re tracking RPO’s or something, you can math it and use regression. But there are players that transcend the system a little bit, like Mahomes will never grade as well as he should. It is a problem completion percentage over expected has too. I think grades are great for many positions, but a big part of betting is learning to up and down weight things, but you’re not going to win taking these things full sale.
NFL Game Breakdowns:
Kansas City vs Jacksonville (37:04)
Buffalo vs Cincinnati (44:07)
Philadelphia vs NY Giants (48:59)
San Francisco vs Dallas (51:38)