How to win the Super Bowl? Louisville company has it down to a science
Louisville data science company Edj Analytics helps NFL coaches make better decisions to win games. It starts with changing their thinking.
LOUISVILLE, Ky. (WDRB) -- When the New England Patriots were down 28-3 to the Atlanta Falcons in the third quarter of last year’s Super Bowl, their statistical chance of pulling out a win was barely greater than 0.
That might have seemed obvious to any football fan.
But Louisville data science company EdjAnalytics actually quantifies such things as part of its business helping a handful of NFL coaches make better decisions.
Edj, a nearly five-year-old startup, runs billions of simulations each year in a proprietary model, all with the goal of helping coaches understand which type of play, whether in the opening series or final drive, increases their odds of winning the game.
Edj co-founders Frank Frigo and Sean O’Leary, along with sales manager Zack Pennington, explained how the company approaches the work in a public talk last week called “How to win the Super Bowl.”
It starts with encouraging coaches to change the frame in which they think about the game.
Do you go for it on fourth down, or do you punt? Most coaches would assess their probability of making a first down.
But for Edj, the right question is, which play increases the chance of winning the game?
On fourth and short, “sometimes you break for a long gain,” Frigo said. “You should be accounting for that.”
The NFL doesn’t allow teams to draw on predictive models to forecast the right play during games, so Edj’s challenge is to train coaches to recognize which behaviors – including common mistakes -- affect their probability of winning in critical situations.
The data crunching leads to some counter-intuitive insights.
For starters, most coaches are too risk averse, O’Leary said.
New England might not be in Sunday’s Super Bowl against the Philadelphia Eagles had Jacksonville “gone for the jugular” on a fourth-and-1 play in the final quarter of the AFC championship game, when the Jaguars were up by 10 points, Frigo said.
Jacksonville punted, fearing the risk of giving the Patriots the ball on the Jaguars' side of the field. But New England drove and scored anyway, narrowing the Jags’ lead to 3.
For Edj, coaches repeat their self-defeating risk aversion in predictable ways.
They often settle for a chip-shot field goal after long drives into the other team’s red zone, Frigo said, when going for it and keeping a 7-point touchdown in sight is the more likely path to a win.
“They say, ‘We just marched the ball 70 yards and we’ve got to get something for this; we are going to put points on the board,” Frigo said. “That is not the best decision in a win-probability context.”
Of course, Edj’s model adjusts probabilities to account for how a “decaying clock” affects win chances. Three points at the beginning of the game might not have the same utility as at the end, depending on the score.
The Patriots' GWC (game-winning chance) during each play of the 2017 Super Bowl:
Another typical mistake: When there are only inches to gain for first down, teams often telegraph a quarterback sneak or short run by lining up in goal-line formation. Their chance of winning the game is better if they run a regular play and expose themselves to all the normal outcomes, Frigo said.
In overtime, how many NFL coaches would have the guts to go for it on fourth down and 3 yards to gain, inside their own 10-yard line? All of them would punt, O’Leary said.
But the reality is that punting is the wrong call, he said, because it too easily puts the opposing team in range for a game-winning field goal. Edj’s data has “proven” this, he said.
“They’re gonna say, ‘If I don’t get it, I lose,’” he said. “You know what, if you punt it, you lose. Eight yards gained, they are in field goal range.”
Edj’s model assigns each team a pre-game probability of winning based on their performance over the season, and the prediction gets better as more games are played, generating more information.
The company wouldn’t reveal whether the Eagles or the Patriots have the edge going into the Super Bowl. Betting markets have the Patriots as a 4-point favorite.
But if the model is right, O’Leary said, “It’s going to be a close game.”