Published: 26 October 2014
As an Eagles fan, I've criticized Chip Kelly a few times for not bringing his propensity for going for it on fourth down to the NFL. At Oregon, he would rarely punt and set the standard for analytical play-calling. While he has had tremendous success already in the NFL, Kelly has been a lot less "risky" in those fourth down decisions. To be fair, it is not the same environment as when he was at Oregon. The job security is not the same. The talent is not the same. The kickers and punters are not the same.
On Sunday, the Eagles had a few tough decisions on fourth down - situations where Kelly would have almost always gone for it in college. First, there was a 4th-and-3 from midfield in the third quarter, down seven, where the expected points numbers say to go for it (+0.11) versus punting (-0.26) - the win probabilities are even closer (25% versus 24%). Kelly punted.
On the very next drive, still down seven in the third quarter, the Eagles faced a 4th-and-2 from no-man's land at the Arizona 36. Kelly sent Cody Parkey out who made the 54-yard field goal. Again, the baseline numbers say to go for it (+1.14 EP, 28% WP) rather than kick a field goal (+0.91 EP, 26% WP). Not a huge difference, so any number of factors could have swayed Kelly's decision toward the field goal. The field goal conversion rate from here is roughly 60%, which is significantly higher than in college.
But, the biggest decision came at one of the most crucial moments of the game. Tied at 17, just after the two-minute warning, the Eagles faced a 4th-and-Goal from the 1-yard line. The announcers were sure Kelly would go for it on fourth down after LeSean McCoy was stuffed on third down.
4th-and-Goal from the 1-yard line converts roughly 52% of the time when going for it. With a touchdown, the Eagles would increase their win probability to 92.19%. A successful field goal means an 80.25% to win the game - and field goal conversion is essentially 100%. An unsuccessful fourth down attempt would result in a 44.36% win probability for the Eagles, as the Cardinals would have the ball in a tie game, but backed up against their own goal line.
E[WPGo-For-It] = 0.52 * 0.9219 + 0.48 * 0.4436 = 69.23%
E[WPFG] = 80.25%
Kelly would have needed to believe there was at least a 75% chance of converting to go for it in that situation. He made the correct call, despite the Eagles ultimately losing. In a game that came down to the last snap, at the last second, every fraction of win probability matters.
Edit: It was actually technically a 4th-and-1 from the 2-yard line, although if I remember correctly, the first down would have been inside about four inches of the goal line. This does not affect the numbers significantly and the only difference is that if the Eagles were somehow able to get a first down instead of a touchdown on a conversion, they could run more time off the clock. That being said, the Cardinals still had all three timeouts remaining, so the Eagles would not be able to burn much time at all.
Keith Goldner is the Chief Analyst at numberFire.com - The leading fantasy sports analytics platform - and creator of Drive-By Football. Follow him on twitter @drivebyfootball or check out numberFire on Facebook. Check out numberFire's new iOS App in the app store now.