Advanced analytics have been a part of various sports for quiet a few years now – baseball of course has been the “pioneer”, with baskeball following up during the past years. For example, many have attributed the increase in the three-point shots to the analytics movement that has revealed that three-point shots (particularly the corner ones) are among the most efficient shots to take (trailing only to shots at the rim). On the contrary, American football has been lagging behind in the advanced analytics move. However, there are two specific play call situations this year that might be strong evidence that this is actually changing…! These two play call situations include (a) the PAT 2-point conversion attempts (which has been the discussion for the past week at the aftermath of the PIT-DAL 6 2-point conversion attempts and the decision of Pete Carroll to go for 2 at the end of the game) and (b) going for it on fourth down. As this blogger has argued both in this blog (here and here) as well as in his academic research, teams are more reluctant in going for the fourth down as they should, while the 2PT conversion should be the default PAT play with the extra point kick being the situational football play (especially after the recent changes at the PAT kick – just check this weeks attempts with 10+ missed extra point kicks!!). Of course, these have not been the way teams approach the game – in fact it is not anywhere close! However, some observations from the first 10 weeks seem to imply that things are changing, and maybe analytics are finally making their way in NFL.
Typically the two-point conversions have been a play that has been called towards the end of a game, either to get a “comfortable” lead or to cover the point differential as the clock is winding down. Therefore, teams typically did not go for 2 in the first half or even in the third quarter. So what I did is to examine the percentage of touchdowns during the first three quarters of the game that are followed by a 2PT conversion. During the first 10 weeks of this regular season 8.6% of these TDs are followed by a 2PT conversion. During the past 7 NFL seasons (2009-2015), for the same first 10 weeks and the same period within the game, on average only 5% of the TDs were followed by a 2PT conversion. Furthermore, this 3.6% difference is statistically significant (at the 0.001 significance level).
Similarly teams are hesitant to go for it on 4th down unless if it is at the end of the game and the team is trailing, even though analytics predict more gains by going for it. This year though again something seems to be changing. I calculated again for the 10 first weeks of the past 8 seasons (including the current one) the number of times that teams have gone for it on a 4th down during the first three quarters. During the past 7 seasons there were on average 61.7 attempts to go for it, while this year there have been 72 attempts (again the difference is statistically significant at the 0.001 level). However, not only that, the average yards to go for the previous years have been less than 2, while this season the average yards to go when attempting a 4th down conversion is 2.4 yards (again statistically significant)! Furthermore, the distance to the goal for these attempts is smaller. This season the average distance to the goal line when a conversion is attempted is 32 yards, while the past 7 seasons it has been 36. Therefore, even though a more comfortable field goal is at stake (and this is the field goal range for the majority of the NFL kickers), teams decide to follow the math advice!
Both these observations might be just a fluke (speaking of p-values and statistical significance) and certainly we need to observe team’s behavior more closely. However, they clearly seem to imply that the analytics are making their way into the NFL and with more detailed data gathered through player sensors this influence of analytics can only grow larger!