clock menu more-arrow no yes mobile

Filed under:

NFL Draft 2015: The Philadelphia Eagles and analytics - how athleticism metrics & positional prototypes play into evaluation and roster building

Note to self: remember to insert a terrible "birds of a feather" pun here

3 Sigma Athlete
3 Sigma Athlete
Howard Smith-USA TODAY Sports

Chip Kelly recently pulled off somewhat of a coup in Philadelphia, taking control of the front office shortly after the 2014 regular season ended. With former GM Howie Roseman re-assigned to deal more with contracts and cap, Kelly promoted the 31-year-old Ed Marynowitz to VP of player personnel. The excellent Sheil Kapadia profiled Marynowitz Thursday, and much of what he said echoed the philosophy that Seattle's built under Pete Carroll and John Schneider. I recommend reading Kapadia's excellent article before continuing as I'm only going to highlight a few passages here. As many on twitter have noted, here's what Marynowitz said:

This is a size/speed league. [Nick Saban and his staff at Alabama] believed the SEC was a size/speed league. There's enough statistical data that will support that in terms of players that are playing at a high level. There's a certain prototype.

A large portion of the best players in the league are also great athletes. As I detailed a few weeks ago, the four current 3 sigma athletes in the NFL are all excellent players. This is anecdotal evidence, but the idea holds up under more rigorous scrutiny. There are not many elite non-QBs who test poorly.

Marynowitz worked for Saban at Alabama before coming to Philadelphia in 2011, and he details their philosophy below.

So there's a certain prototype at each position. We try to build the same thing here, whether it's at inside linebacker, outside linebacker, corner, safety. There's a prototype, and there's a model that fits what we do.

I've written before about the Seahawks and prototypes, an application that I refer to as "roster mirroring." If the backups on a team fit into the same athletic class and build as the starters, then the scheme used to maximize the strengths of the starters should also serve the players waiting on the bench. Next Man Up works much better when the incoming player resembles the starter they're replacing.

Marynowitz also spoke on a general height/weight/speed model and trimming the draft pool of potential Eagles, summing it up with the following quote:

If you have seven draft picks, do you really want to waste one, especially in the top three rounds, on a guy that history is telling you... typically these guys with these types of measurables don't produce at this level?

As Chris Brown tweeted, this is an application of Bayesian reasoning, which you can read about here.

The application of athletic comparisons is frequently debated, but this quote does a good job of relating how they should be used. There are between 2,000 and 3,000 players who test at either the NFL Combine or pro days during the pre-draft season. The idea is to take the best possible bet, and if a certain prospect has athletic comparisons that show significant issues at the NFL level, I'd prefer to move on to the next player.

We don't have to be right about every player. Missing on a prospect who ends up overcoming the odds is frustrating, but it happens. I'm more concerned about not being wrong, and that means thinning the pool of possible selections. Kapadia reports that the Eagles only have around 150 players on their final draft board, summing it thusly:

The specific approach can shrink the pool of potential prospects, but Kelly, Marynowitz and others find more danger in trying to gamble on exceptions.

Pete Carroll has spoken before along the lines of "if you keep making exceptions, your whole team ends up being made of exceptions." Notably, Seattle and Philadelphia had 2014's two highest-ranked SPARQ rosters by my pSPARQ calculations.

This is less about individual evaluations and more about macro-level team roster strategy. As I wrote in my first post on 3SA, a large group of superior athletes will ultimately perform better than a large group of averages athletes. Analytics matter most when used over a large number of decisions, the small percentages adding up to meaningful and significant added value. By generally selecting players from an athletic pool and taking fewer shots on the exceptions, it's possible to obtain more value in the long run.

Marynowitz did note that exceptions aren't necessarily discarded from the board, but that they "better be exceptional in a lot of other areas" to be considered.

Kapadia goes on to interview Marynowitz about how the team views scheme fit and culture in prospect evaluation. It's excellent and well worth taking the time to read if you ignored my advice at the start of this article.

The article makes clear that analytics don't make every decision in the Philadelpha front office; still, the Eagles do subscribe to an analytical ideology which plays a significant role in shaping their draft board and roster composition.


I'll be back next week with the final SPARQ update and a few articles searching out roster mirroring candidates for Seattle in the upcoming draft.