THE CLAIM
The New York Giants suffered a 24-21 Thursday night loss to the Washington Commanders, a game that potentially turned on a handful of fourth-down choices. Up four with 3:27 left, John Harbaugh’s Giants had fourth-and-1 at their own 11. The Giants lined up in a run formation, ran a comical hard count to “Who Let the Dogs Out,” took a delay of game, and punted. It would have been an aggressive call to go for it by conventional standards, but a no-brainer by the numbers.
The Commanders were intercepted on the subsequent drive, but you can argue that the decision gave the Commanders a needed additional drive to subsequently get their game-winning touchdown.
For years one of the league’s more aggressive fourth-down coaches with the Baltimore Ravens, Harbaugh has gotten analytically cold feet with the Giants, but has it cost them wins?
THE VERDICT: LEAN CONFIRM
Harbaugh’s conservative decisions are costing them, especially on Thursday night. The Giants have gone for it on only 4 of 13 fourth downs where the nfl4th model shows that going adds at least 1% in win probability. That 30.8% rate makes the Giants 29th in the NFL, against a substantially higher average of 46.7%. On the clearer recommendations, a 2% edge or more, they have gone 4 of 10 chances, which ranks 28th. Looking across all fourth-down opportunities, the Giants failed to go for it 18 times that the model showed positive win probability, for 28.2% total, or 3.1% per game. That’s 6th-most of any team in the NFL, against a league average of 2.5%.
In the Commanders loss alone, four decisions left 13.5% in win probability on the table, with 12.0% in clear go-for contexts (+2% WP to gain). In a close game with a three-point margin, that is enough win probability lost to plausibly flip the result.
The season-long pattern for the Giants is conservative, a change from both Harbaugh’s Ravens teams and last year’s Giants staff under Brian Daboll.
THE EVIDENCE
The costliest call on Thursday was not even the one that went viral. On fourth-and-3 at the Commanders’ 46 with 14:15 left in the fourth quarter, down 3, the model said go by 6.3%, with a 48% chance of a first down. The Giants punted into the end zone for a touchback. That was the Giants’ biggest fourth-down-decision win probability loss of the season.
The hard count is the cleaner conservative choice. Before the penalty, fourth-and-1 at their own 11 was a go by 3.3%, with a 67.4% chance of converting. They took a 4-yard delay — the hard count did not draw offsides — and it became fourth-and-5 at their own 7, where punting was then clearly the right call.
The two other spotlighted calls did not project as egregious decisions by the model. On fourth-and-7 at the Commanders’ 37 in the first quarter, the Giants punted, with the alternative being a 54- or 55-yard try for rookie Dominic Zvada. The model preferred going for it by 1.4%, with the field goal and punt rated about even. For context, Zvada is 1 of 3 from 55-plus this year. And on fourth-and-5 at the Commanders’ 41 with 1:54 left, up four, the model agreed with the punt decision, by 1% over going. The problem was the punt execution, as Jordan Stout shanked it 12 yards.
The Giants did properly go for it once on fourth-and-1 at the Commanders’ 30 in the second quarter and converted.
Thursday continues a season-long pattern for the Giants: they punted on fourth-and-1 at their own 29 (3.4% lost) and fourth-and-5 at their own 46 (2.2% lost) in a win at the Philadelphia Eagles, on fourth-and-13 from the Houston Texans’ 37 when a 54-yard field goal would have been better by 4.7%, and on fourth-and-2 at their own 32 (2.3% lost) in the Week 5 tie with the Commanders. Harbaugh’s four correct +1% WP decisions to go were all fourth-and-short near midfield or in plus territory.
