Jesus, people. From what I can gather - there is no "statistical" advantage to deferring. If so, why do teams defer 75% of the time? Coaches believe there is a momentum opportunity by being able to score on two straight drives. Or Belichick started doing it, so everyone else did.
From here:
https://www.numberfire.com/nfl/news/7416/should-nfl-teams-defer-after-winning-the-coin-toss
The article also mentions that deferring teams have been more successful that those receiving, but Mays does much better job putting the results in context than
some others who claim that the higher winning percentage of deferring teams actually proves anything. The reason the higher winning percentage doesn't mean anything is because the teams choosing to defer is not a random sample. It could just be that better teams -- like the Patriots -- are choosing to defer more often, and that's skewing the results.
Something absent in the many articles -- and I was shocked by this -- on deferring is a look back at pre-deferral winning percentages, where the kick and receive cohorts were randomly assigned. Remember, before a team could choose to defer, winning the coin toss meant you had to receive, or else you where giving your opponent the ball first and the choice to receive in the second half -- not a bright move.
Well, we have that data to analyze. I took pre-deferral data (2000-2007) and calculated the winning percentages for those who lost the coin toss and kicked to start the game (exactly the same as those choosing to defer now), and those who won the coin toss and received the opening kickoff.
| Win Percentage |
---|
Kicking Team | 48.8% |
Receiving Team | 51.2% |
The results over a sample of 2,110 games, with teams randomly assigned to the kicking and receiving groups, show that there probably isn't any advantage to kicking first and receiving in the second half. In fact, the results show that there could be an advantage to receiving first; however, a binomial test on the results yields a p-value of 0.13, or not statistically significant at the commonly used 95 percent confidence threshold.