Frank Frigo

It struck me that those kinds of decisions are very, very similar to backgammon decisions and decisions that were modeled

One of the great breakthroughs in the modeling of games, in backgammon in particular, was that when you focus on the proper objective, which in backgammon is giving yourself the highest probability of winning the match.

And in football, if you apply the same metric, which is

How do I make a decision that gives me the highest expected outcome of a win, not maximizing yardage, not assuring I get some amount of points on a particular drive?

When you shift that objective and that metric, it really opens things up, and you start to see that...

These risk-averse decisions are actually quite wrong.

We set out to build a model that could actually measure it.

We did this by building a fully customizable simulation model that was built on a lot of NFL empirical data.

My partner, Chuck Bauer, who is also an experimental physicist who did a lot of research at Indiana University and worked for NASA, he and I put our heads together to build a model that could simulate NFL games from start to finish, running clock, penalties, the whole thing.

We had a hypothesis that coaches were giving up quite a bit.

It turned out after we started studying it more deeply that they were worse at it than we even thought they were.

They were giving up even more.

Then we got excited.

We were like, it's time to start knocking on some doors.

I would say it influenced probably just about every team, some more directly, some indirectly.

We licensed to about 12 NFL teams.

Most famously, we worked with the Philadelphia Eagles during their Super Bowl run.

They used it religiously.

We had buy-in from top to bottom, from Jeff Lurie, the owner, to Howie Roseman, the GM.

to Doug Peterson, and also with the Kansas City Chiefs.