Sabermetrics in Drafting Football Players

**Pavlovs Dog** · 02-25-2015, 12:00 PM

Hello, I'm a guy who has been reading this site for years, but I don't post much because I learn so much more from listening than from speaking. But I've got a question.

The book (and film) Moneyball shined a bright light on sabermetrics, or the use of statistics to predict player performance. I can't help but imagine, given the huge investment that NFL teams make in the draft, that they don't employ statistical analysis in order to help their drafting. I realize that it is both art and science, but it would seem that as a GM, you would want to get the science part right.

I'm certain that NFL teams must have calculated correlation coefficients for college players in an attempt to find the relationships between NFL success and variables such as 40 yd dash times, cone drill times, bench presses, Wonderlic scores, college statistics, captaincy, competition levels, work ethic, academic achievement, hand size, etc.

Presumably, these variables would have different success coefficients for different positions (e.g. 40 yd dash speed is more important for WR and CB). Please forgive a long and nerdy question, but does anyone know how seriously NFL teams track this data? One never reads about this subject in the press or on discussion boards.

**Kyle Cactus** · 02-25-2015, 12:36 PM

I think part of the reason that there isn't much to read on this is that NFL teams are pretty secretive.

Another part is that I still think most NFL teams are only dipping their toe in the water as far as analytics are concerned. Granted, that's speculation on my part.

There's also a culture in football sometimes that's resistant to change. I'm sure that if analytical methods prove their worth, they will eventually catch on around the league. It will just take time.

Here's a recent thread on the subject.

**wickedsolo** · 02-25-2015, 12:41 PM

I'd imagine that every team has a baseline metric for grading players based on their position.

For example, some teams out there may not consider a receiver that runs higher than a 4.65 40-yard dash. Drilling down, that same team may rule out receivers that are sub-6' and sub-200lbs that cannot run under a 4.60 40-yard dash, but they'd be ok with a 4.60 receiver if they're 6'1" 201lbs on up.

There are SO many different metrics and combinations of statistical measurements that could be applied.

Then, there are outliers like an Anquan Boldin who ran something like a high 4.7/low 4.8 at the combine and have had borderline HOF careers.

I just don't see how it's possible to apply saber metrics to functionally predicting a draft pick's potential and so on.

I think that it could be part of a larger puzzle though.

**DC Raven** · 02-25-2015, 12:50 PM

The Ravens are a great example, actually. The Team employs two really well credentialed Analysts for statistical analysis:

Sandy Weil, a Yale and Carnegie Mellon graduate with Wall Street experience, has been the team's director of analytics since 2012. Eugene Shen, a Harvard and MIT grad with years of experience in applying advanced metrics to the financial markets, works directly with the coaching staff.

I believe they do most of their work on strategic analysis, such as the probabilities around electing to receive or defer at the coin flip, going for it on 4th down, and so on. But, I find it hard to imagine that the team pays these two people 6 figures (people with those credentials easily command 200k+) for just that. I'd be willing to bet they dabble in exactly the types of draft analysis you mentioned: correlations of WR 40 times with receiving yards during their 4-year rookie contract, for instance, would be a really simple/crude but apt example.

As someone who has a finance background, and by extension, some cursory knowledge of neat statistics stuff - I can vouch that there are some really interesting models they could produce - and I hope they are. Numbers don't lie! I think they would find that 40 times, for example, are mostly garbage as an independent measure of ability and production. In any case, this is really cool stuff and I'm glad you made the thread. Here's to hoping the Ravens continue to develop their use of analytics, particularly in drafting.

**Greg** · 02-25-2015, 09:49 PM

Originally Posted by DC Raven

The Ravens are a great example, actually. The Team employs two really well credentialed Analysts for statistical analysis:

I believe they do most of their work on strategic analysis, such as the probabilities around electing to receive or defer at the coin flip, going for it on 4th down, and so on. But, I find it hard to imagine that the team pays these two people 6 figures (people with those credentials easily command 200k+) for just that. I'd be willing to bet they dabble in exactly the types of draft analysis you mentioned: correlations of WR 40 times with receiving yards during their 4-year rookie contract, for instance, would be a really simple/crude but apt example.

As someone who has a finance background, and by extension, some cursory knowledge of neat statistics stuff - I can vouch that there are some really interesting models they could produce - and I hope they are. Numbers don't lie! I think they would find that 40 times, for example, are mostly garbage as an independent measure of ability and production. In any case, this is really cool stuff and I'm glad you made the thread. Here's to hoping the Ravens continue to develop their use of analytics, particularly in drafting.

This. I imagine they amass a ton of data but put little to no weight on bench presses, 40 times, and the like. My guess is they gauge made and missed tackles, percentage caught of catchable balls, precision of routes run, blocks made (much like FilmStudy does), penalties, coverage and ball skills of DBs, etc. I would guess they put their eye on real plays made and not track star statistics.

**Bigfish** · 02-25-2015, 10:14 PM

Originally Posted by DC Raven

The Ravens are a great example, actually. The Team employs two really well credentialed Analysts for statistical analysis:

I believe they do most of their work on strategic analysis, such as the probabilities around electing to receive or defer at the coin flip, going for it on 4th down, and so on. But, I find it hard to imagine that the team pays these two people 6 figures (people with those credentials easily command 200k+) for just that. I'd be willing to bet they dabble in exactly the types of draft analysis you mentioned: correlations of WR 40 times with receiving yards during their 4-year rookie contract, for instance, would be a really simple/crude but apt example.

As someone who has a finance background, and by extension, some cursory knowledge of neat statistics stuff - I can vouch that there are some really interesting models they could produce - and I hope they are. Numbers don't lie! I think they would find that 40 times, for example, are mostly garbage as an independent measure of ability and production. In any case, this is really cool stuff and I'm glad you made the thread. Here's to hoping the Ravens continue to develop their use of analytics, particularly in drafting.

Never realized they had eggheads like this on the staff. I'm sure they have some advanced analytics that would blow our minds. Having worked with analyst on risk models in the past, it's a bunch of data fields assigned weight and run through some crazy algorithm. It's constantly scrutinized and calibrated. In the end, I don't think it's a end all, but it may tip the scales when things are close.

**Pavlovs Dog** · 02-26-2015, 11:13 AM

Thanks to all for some great and thoughtful answers. Specifically, it would seem to me that statisticians would be interested in attempting to quantify the strongest factors to predict boom or bust. Why do some players (Tom Brady, Matt Birk, Shannon Sharpe) who are drafted late go on to become bona fide stars for years, while others who are drafted early (Ryan Leaf, et. al.) become total busts?

In retrospect, some "experts" who were suspiciously absent in advance want to point out some invented "obvious" factor ("he has heart" or "he was all hype"), but I can't help but thinking that these traits were predictable in advance, yet still ignored by professionals.

How does a team factor in the guy who was a star in college, but played in a lesser conference where the competition was weaker? Brandon Williams, Lardarius Webb, and Joe Flacco all come to mind. I would venture that, for these players, performance against top-flight competition at events like the Senior Bowl must be critical.

It's surprising to me that even Ozzie Newsome, who is widely acknowledged as being among the best of talent evaluators, still misses just about as often as he hits on draft picks. Obviously, injuries are a factor (Sergio Kindle, Dan Cody), but a lot of players just aren't good enough.

**WrongBaldy** · 02-26-2015, 05:35 PM

Adam Schefter @AdamSchefter · Jags Sr. VP Tony Khan bought 100% of TruMedia, an engineering firm used for data analytics by ESPN + NFL/MLB/soccer teams. Analytics growing

**Trackmaster** · 03-17-2015, 08:47 AM

Sabermetics, as it gets its name from (Society for Baseball Research) is a match made in heaven for baseball, as most of your production is from digital, cut and dry, perfect world scenarios (the pitcher vs. the batter where there are only so many outcomes). A far distant second in the race to use advanced metrics is basketball. Most of the other sports are tough because everything is so team oriented, and not as turn based as baseball. Football is turn based, but is the ultimate team sport, lending itself to more evaluation from individualized combines and tests.

At the heart of it, sabermetrics is making decisions based on performance, using some MIT algorithms, and trying to take luck out of it. The NFL will get there, but you have to have the right people making decisions, and use the right stats. The book Moneyball talks about stats that are just a bunch of numbers vs. language. You want analysis that helps you, and makes things cleaner and clearer. Not a bunch of pseudo science that helps nobody and embarrasses the whole field (most hits on a Tuesday afternoon game in the rain). Maybe a stat that can accurate predict how fast a given player will recover from an ACL injury, or the likelihood that a 17 year old Dominican kid with 100 walks per season and gold glove potential will develop into a 30 HR per season slugger.

**BcRaven** · 03-17-2015, 12:47 PM

To balance out the intelligentsia, the Ravens also employ a fat guy (ALA Art Donovan) who looks over the FAs and prospective draftees and says, "Yeah that guy's OK, he bought me a couple hot dogs." OR "Don't touch this guy, looks like he hasn't had a meal in a month." There are 'Sabermetrics' and 'Sausage and Brew' science that when blended carefully, lead to astute decisions... Bc

**Tenacious Faulker** · 03-17-2015, 01:30 PM

Clearly stats are used more and more to discern player success in football. Sacks weren't documented until the 80's and tackle weren't until the early 2000's. Offensive linemen are now getting graded better than the number of games played and sacks allowed with things like QB harries/hit, pass/run blocking efficiency. Since Deone Sanders cornerback have been getting measured in number balls thrown in their direction and opposing QBR.

Yet, for years, teams scouts knew inherently, if not precisely, how a team could benefit with a certain player. Teams have players to fill certain roles - always have, always will. As the sport grew ricjer and more competitive teams began hiring professional statisticians and actuarialist to come up with their own internal, proprietary modelling.

Cybermetrics is just one form. If it's in use in some form in the NFL I can;t say for sure, but there are probably 32 varying models in the NFL used in irecting teams' risk vs. reward vs role for players.

**Shas** · 03-17-2015, 01:59 PM

Originally Posted by Trackmaster

Sabermetics, as it gets its name from (Society for Baseball Research) is a match made in heaven for baseball, as most of your production is from digital, cut and dry, perfect world scenarios (the pitcher vs. the batter where there are only so many outcomes). ...Most of the other sports are tough because everything is so team oriented, and not as turn based as baseball. Football is turn based, .

This is along the lines of my reaction to the question.

The OP was thinking about 40 times and bench presses, which is not a direct measure of in-game performance. Game performance stats are in the DNA of baseball, but not for a sport like football, aside from some macro-level stats for skill players.

That said, you can look at what Football Outsiders does with individual, in-game stats. They even take the step of normalizing the data. Meaning, they will take into account quality of opponent, down, distance, score, etc. That's about as close as you can get to Sabermetrics.