Down by three points with 1:07 to play and one timeout left in Sunday's divisional-round game, Dallas Cowboys quarterback Dak Prescott completed an 11-yard pass to Jason Witten for a first down on the Green Bay Packers' 40.
The Cowboys hurried to the line and spiked the ball, stopping the clock with 48 seconds left, saving their final timeout. Spiking the ball in that situation is not the usual tactic, and was very likely a mistake, but it only seems so costly because of the improbable events that followed in the Packers' thrilling 34-31 win. After the Cowboys advanced the ball only 7 yards and kicked a 52-yard field goal to tie it, the Packers marched down the field on their ensuing possession, setting up Mason Crosby's 51-yard game-winning kick.
Spiking the ball, of course, exchanges one down for about 10 or 15 additional seconds on the game clock. It was an unorthodox tactic in this situation because the Cowboys still had a timeout and were not yet in field goal range. Spiking the ball is also typically done when teams intend (or need) to use their fourth down to try to convert, which still allows them three plays to try to pick up 10 yards rather than just two.
Because of the game situation -- down by three with time dwindling -- Cowboys coach Jason Garrett needed his fourth down to attempt a field goal to tie, which left his offense with only two chances to convert. The 40 is only desperation field goal range, and so Dallas needed all three downs to try to get closer.
On the other hand, preserving the timeout would have been the right thing to do had the Cowboys converted the first down. It keeps the middle of the field open as an option for Prescott -- or possibly even an Ezekiel Elliott run -- and it allows the field goal team to set up for a kick on a subsequent set of downs. Normally a field goal unit can set up and kick well within a 40-second play clock, but that was no longer the binding constraint. The game clock would have been at less than 40 seconds on the next set of downs, and it might have expired on Dallas even if it had a spare down to spike the ball. And of course, saving the timeout also helps Dallas should it try to win the game with a touchdown rather than go for the tie.
We can weigh these considerations in our head all we like without really knowing how smart or how costly the decision to spike the ball was. Fortunately there are methods to measure hard questions just like this. A game simulation might be the best tool here because it captures the nuances of timeout values and does a slightly better job than our win probability model valuing downs other than first and fourth down.
Here's how this works: We tell the simulation to start at a certain state of the game -- score, time, down, timeouts left, etc. -- and it will simulate the rest of the game (using actual historical play outcomes from similar situations) thousands of times. The proportion of times that the offense wins in those simulated games is our estimate of win probability.
Spiking the ball
This decision left Dallas with a second-and-10 from the Green Bay 40 with 48 seconds left and one timeout. The Packers had two timeouts. With this combination of inputs, Dallas wins 32 percent of the time by spiking.
Using the timeout
In this scenario, Dallas would have a first-and-10 from the 40 with closer to 55 seconds and no timeouts left, giving Dallas a 35 percent chance of winning.
Letting the clock run and running a play
Assuming the clock runs an additional 15 seconds, that would give the Cowboys a first-and-10 from the 40 with 33 seconds and one timeout left. That's about the same as the spike, giving Dallas a 32 percent chance to win.
So using the timeout would have been the best thing to do in a typical game. Spiking the ball doesn't buy Dallas any advantage, yielding approximately the same chance of winning as simply running another play. But the cost of the error is relatively small -- just a 3 percent drop in the Cowboys' chance of winning.
It's just that everything that happened after the spike broke against Dallas, magnifying the cost of the error. The Cowboys failed to convert the first down. They didn't use any time in the process, leaving the Packers enough time to drive for their own field goal. And of course, Aaron Rodgers was amazing, setting up Crosby to make a second 50-plus-yard field goal in the final minutes to win.
Decisions have to be judged based on the information at hand at the time. Although that sequence of outcomes was certainly possible -- especially with Rodgers on the other sideline -- it was highly improbable, and there's no way Garrett could foresee exactly how things would unfold. That's why the spike wasn't as costly as we might think.
Even if it was a mistake, I would credit Garrett with at least having the right mentality. Saving the timeout would have been the best decision if Dallas intended to win the game with a touchdown rather than play for the tie. All too often (and Garrett has also been guilty of this) coaches play for a long field goal to tie the game rather than try to win or at least get closer for a higher-probability kick.
Things didn't work out for Garrett's Cowboys this time, but his overall strategy was the right one.