Last week, I utilized a Win Probability model to evaluate a key decision made by Jason Garrett against the Oakland Raiders. According to the model, Garrett made the right call to kick a late field goal rather than go for a touchdown on 4th and goal from the 1 yard line.
Sorry, fans, you have to put the pitch forks away for at least that case.
This week, we go back a bit further to the week four match up against the Los Angeles Rams. At 2-1 Dallas was looking to begin to build early season momentum and get a key conference win at home. Unfortunately, the Rams climbed back into the game in the 2nd half and defeated the Cowboys, sending their seasons in two very different directions.
The specific decision we will evaluate came with 5:11 left in the 3rd quarter, with the Cowboys holding a slight 24-19 lead. Here the Cowboys faced a 4th down and 2 from their own 47 yard line. Effectively at mid-field and needing an offensive spark, all of Cowboys Nation let out a collective sigh when Garrett decided to punt the ball away rather than go for the first down.
After all, you have the best running back in football, right?
The Cowboys punted it away to the Rams, and when they got the ball back they had lost their lead. A lead they would never regain, by the way. But what does the model say about the Cowboys' decision?
According to the model, the Cowboys had about a 79% chance of winning the game prior to making any decision. After punting the ball away and giving the Rams the ball at their own 18 yard line, that probability really didn't move much at all. In fact, it only moved positively by about .6%.
Had the Cowboys gone for it and got the first down, let's say by gaining 3 yards and getting to midfield, Dallas would have then had an 87% chance of winning the game. Had they failed in their attempt and given the Rams the ball back at the 47 yard line, that probability would have fallen from 79% pre-decision to 70% post-decision.
Of course, you can evaluate these numbers in a few ways. The Cowboys' chances of winning actually marginally increased when they punted the ball away, so punting is the safest choice. However, the lower-bound of their win probability had they went for it and failed was still as high as 70%. This is only about a 10% decrease from where it is when they punt the ball back to the Rams.
A risk-averse coach would punt the ball and “live to see another play.” But, I would say that the smart decision here, based on the numbers, would have been to go for the first down. The Cowboys' offense had had success against the Rams much of the afternoon, and even in a world where the worst case scenario occurs, you are still heavy favorites to win the game.
