It was more than a decade ago when Bill Belichick made a then very controversial decision to go for it on 4th and 2 from his own 28-yard line with a 6-point lead against the Indianapolis Colts. At the time, we defended his choice as both courageous and analytically sound. https://fifthdown.blogs.nytimes.com/2009/11/25/the-anatomy-of-a-fourth-down-decision/
Although Belichick took considerable heat after coming up short and handing the ball back to Peyton Manning for the Colts’ victory, this decision illustrated just how far ahead of the curve he was at that time.
Going for it on 4th and short in your own territory with a lead may still be controversial in 2020, but it certainly isn’t as big of an outlier as it was in 2009. Analytics, and particularly the merits of highly leveraged 4th down aggression, is beginning to change the NFL. Belichick has always been considered one of the most innovative coaches relative to his peers, but how is he keeping up? As Aaron Schatz of Football Outsiders sums it up, “the problem isn't as much that Belichick is less aggressive now on fourth downs, but more that the league as a whole has passed him by. If you look at the period from 1986 through 2017, Belichick comes out as one of the ten most aggressive coaches on fourth downs (minimum three years). But as the league as a whole has started going for it much more often on fourth downs, Belichick hasn't changed, so his rank in the league has dropped to average or below.” Even while the Patriots were posting impressive results in recent years, we began to notice just how much GWC they were squandering on suboptimal 4th down decisions.
*High Confidence errors are a subset of total errors that pass a stress test of a strong counter argument to the model’s recommendation. For example, if the model indicates a team should have attempted a first down on a 4th and short, the simulation is rerun by substituting a very weak offense and a very strong opposing defense to see if the optimal choice is overturned. Only errors that are upheld by this criterion are labeled “High Confidence”.
While it may appear as though Belichick is improving slightly during this period, the cost of approximately 1.6 expected games in 2017 was an NFL worst. So far in 2020, the Patriots are ranked 20th with regard to total GWC cost. It should be noted that very strong offenses, as was the case with the Patriots in 2017, are typically under greater scrutiny in our analysis because of their higher expected conversion rates. In their Super Bowl loss to the Eagles that season, the Patriots gave up more than 20% GWC on high confidence 4th down errors.
Much of the Patriots’ very disappointing 2-5 record can be attributed to the loss of Tom Brady. The Patriots offense is currently 6th in rushing and 23rd in passing according to the EdjSports Power Index. An example of how Belichick is missing the mark this season occurred during their close loss to the Bills yesterday.
Trailing 14-6 with 7:36 remaining in the 3rd quarter, the Patriots chose to punt on 4th and 6 at the Buffalo 43-yard line. According to our analysis, this decision costs the Patriots -2.2% GWC on average. Jake Bailey managed to pin the Bills on their 8-yard line, but surprisingly, even if this punt was downed at the 1-yard line it would have still shown a slight GWC cost to the Patriots. The average historical starting position for a punt of this type, inclusive of touchbacks, is the 17-yard line. While the actual punt was much better than average, it was still costly to the Patriots’ winning prospects.
With 9 games remaining this season, Bill Belichick will once again need to find ways to generate an advantage against the competition. With the rising tide of analytics adoption among his peers, this won’t be an easy task.