Case Study 2 — Super Bowl LII: Two Fourth Downs and a Reputation

A historical case. The game facts below are widely documented; no quotes are attributed to real people, and the expected-points values are this book's illustrative round numbers, not figures from any published model.


Setup

The game. Super Bowl LII, played in February 2018 in Minneapolis: the Philadelphia Eagles against the New England Patriots. Philadelphia's head coach was Doug Pederson. Its starting quarterback that night was Nick Foles, the backup who had taken over late in the regular season after Carson Wentz was injured. The Eagles won, 41–33.

The reputation. The 2017 Eagles were widely noted for going for it on fourth down more often than most NFL teams, and their staff was widely reported to use analytics in those decisions. That matters to this case for a reason that has nothing to do with spreadsheets: by the Super Bowl, every opponent knew it. A team with a reputation for fourth-down aggression changes how its opponent plays third down, how it manages the clock, and how it thinks about field goals — the same "four-down territory" effect described in 28.6.

The two decisions. Two fourth-and-short decisions in that game are part of NFL history:

  1. Late in the first half, leading 15–12, Philadelphia faced fourth-and-goal at the New England 1 with well under a minute left before halftime.
  2. Midway through the fourth quarter, trailing 33–32, Philadelphia faced fourth-and-1 near its own 45.

This case uses the chapter's tools — the break-even number, "name it, then price it," and the separation of decision, call, and execution — on both. It's told from the outside, the way a coordinator studies another team's choices: we know what happened, not what was said.


What happens

Decision 1: Fourth-and-goal at the 1, end of the first half

The conventional choice is a near-certain field goal: three points, a 15–12 lead becomes 18–12, and you go to the locker room. The alternative is a touchdown try from the 1.

4th-AND-GOAL AT THE 1, UNDER A MINUTE LEFT IN THE HALF
(illustrative values; ignores the kickoff and the clock's tiny tail)

 KICK --> field goal, near certain ............................ ~ +3

 GO ----> score (p)   : touchdown + extra point ............... ~ +7
     \--> fail (1-p)  : opponent ball at its own 1, seconds
                        left, nothing but a kneel ............. ~  0

          GO overall = p x 7 + (1 - p) x 0 = 7p
 Break-even: 7p = 3  -->  p = about 0.43
 Go if you convert from the 1 more than about 43 times in 100.

📊 Diagram (described): A decision tree for fourth-and-goal at the 1 just before halftime. Kicking leads to one outcome: a near-certain field goal worth about three points. Going splits into two: a touchdown and extra point worth about seven, or a failure that leaves the opponent at its own 1 with only seconds left — worth about nothing to either side, because they'll simply run out the half. Going is worth seven times the conversion chance, so it beats the kick whenever you convert more than about 43% of the time.

Two things make this go stronger than a typical fourth-and-goal. First, the failure is unusually cheap: with seconds left before halftime, the opponent pinned at its own 1 has no realistic way to score, so the failed branch is worth roughly zero instead of a negative number. Second, from the 1, conversion rates are generally better than a coin flip in public data across many seasons — the defense has no depth to defend. Both push the true chance comfortably above the break-even.

Philadelphia went. The call was the play now known as the "Philly Special": the ball was snapped directly to running back Corey Clement, who pitched it to tight end Trey Burton, who threw it to Foles — the quarterback — uncovered in the end zone. Touchdown. With the extra point, 22–12 at halftime.

Decision 2: Fourth-and-1 near their own 45, fourth quarter

Late in the game, expected points give way to win probability: the score and the clock matter more than the average next score. Trailing 33–32 midway through the fourth quarter, the Eagles faced fourth-and-1 in their own territory.

The conventional choice is a punt: pin the opponent, trust the defense, get the ball back. But look at what the punt actually buys against that opponent at that moment: New England's offense had been moving the ball most of the night, and a punt hands a one-point lead and the clock to one of the best late-game quarterbacks in league history. A failed fourth down is worse — the opponent starts near field-goal range — but a conversion keeps the ball, the clock, and the chance to take the lead with a drive.

The break-even here is much higher than on the goal line, because the failed branch is costly. But the punt branch is costly too: in that game state, the punt isn't the safe choice; it's a bet that your defense will get a stop it hadn't been reliably getting. Public fourth-down models of the kind described in 28.6 generally favor going on fourth-and-1 near midfield in close fourth-quarter games, and this situation was squarely in that territory.

Philadelphia went. Foles completed a short pass to tight end Zach Ertz for the first down. The Eagles kept the ball and scored the go-ahead touchdown later on that drive, and their defense then forced a fumble on New England's next possession. Final: Eagles 41, Patriots 33.


Analysis

Decision, call, execution — graded separately.

Decision 1 (4th-and-goal at the 1, end of half) Decision 2 (4th-and-1, own territory, Q4)
Decision Strong. Low break-even (~43%) and a cheap failure branch. Defensible to strong. High cost of failure, but the punt was not "safe" in that game state.
Call High-variance by design: a direct snap and a throwback that put the ball in the hands of the player the defense had stopped covering. Low-variance: a short throw to a reliable tight end — a call whose likely failure was an incompletion, not a turnover or sack.
Execution Clean — snap, pitch, throw, catch. Clean.
Result Touchdown First down, and a go-ahead drive

Notice that the calls were very different in character even though both decisions were aggressive. The goal-line call was a gadget, chosen for a situation where the failed branch was nearly free. The fourth- quarter call was a simple, high-percentage throw, chosen for a situation where failure was expensive. That is exactly the relationship this chapter describes: the model prices the decision; the coordinator makes the conversion chance true with a call that fits the cost of failure.

What the result doesn't prove. It is tempting to say "analytics won the Super Bowl." That's resulting in the other direction. If either play had failed, the decisions would have been just as sound — and, almost certainly, far more widely criticized. A fourth-down process should look identical in the week-after review whether the ball crossed the line or not. The best evidence that a staff's fourth-down process is sound is not one famous night; it's a season of decisions, graded before the results are known.

What the reputation did. A team known for going for it changes the opponent's arithmetic on every third down in four-down territory. The defense can't treat third-and-2 near midfield as a single snap; it must defend two. That's an invisible return on a fourth-down philosophy that never shows up in the fourth-down conversion column.

Lyle's view. I come back to this game for one reason: the two calls. Plenty of coaches can be talked into a fourth-down "go" by a chart. The harder skill is having the right kind of call ready for each kind of go: a play you'd risk when failure is cheap, and a play you'd trust when failure is expensive. On our Green panel, those live in different rows. That's not an accident.


Discussion questions

  1. Compute the break-even for Decision 1 if the failure branch were worth −1.0 instead of 0 (for example, if two minutes remained and the opponent could drive). Does the decision change?
  2. Why did the fourth-quarter decision depend on win probability rather than expected points? Name two game-state facts that a win-probability view captures and an expected-points view doesn't.
  3. Compare the character of the two calls. When would you choose a gadget on fourth down, and when would you never choose one?
  4. Suppose the fourth-quarter pass had fallen incomplete and the Patriots had kicked a field goal to win. Write the two-sentence Monday review using "decision, call, execution."
  5. How would an opponent's knowledge of your fourth-down reputation change its third-down calls, and how should you exploit that?

Your turn

Build two rows for your own Green panel: "Fourth-and-short, failure cheap" (end of half, goal line, opponent pinned) and "Fourth-and-short, failure expensive" (own territory, close fourth quarter). Put two calls in each row. For each call, write one sentence on its likely failure (an incompletion? a loss? a sack? a turnover?) and why that failure fits the row. Then compare your two rows: if the same play appears in both, explain why it's safe enough for the expensive row.