If the more than 600 comments on Matt Warren’s Five observations from the New York Giants game is to be believed, people have opinions about the decision by Sean McDermott to take a knee to close out the first half of the game. There’s reasonable arguments to be had on both sides of the fence but what do NFL coaches have to say about it? Or rather, what does the data say they lean toward?
Methodology
If you’re into football analysis one of the best sites you can bookmark is Pro Football Reference. It’s one of the most comprehensive databases for sports fans where you can find just about anything you’re looking for. And if you get to know their play index tools you can create data magic.
Using the play finder let’s see if we can get a handle on how often similar plays occur. To do that we need a lot of facts about the play. The kneel occurred with 47 seconds left in the second quarter. The Buffalo Bills were on their own 28-yard line. They were up by 14 points.
Now, remember, I said similar plays. If we put in the exact criteria, we come up with one play that meets all these parameters since the 2009 season: The one from last Sunday against the Giants. Let’s set the stage for “similar” then:
- We’ll go back to the 2009 season through today to get a more valid sample size
- Starting field position will be set between 18 and 38 yards, or ten yards in either direction of the play we’re looking at
- The score will be set to leading between 7 and 21 points, or 7 on either side of the Bills
- Time is set 12 seconds on either side of the Bills, so between 59 and 35 seconds left in the second quarter
- Plays will be limited to first downs only. This is due to the nature of the discussion at hand. If a team goes for it on first down, doesn’t like the result and then kneels we want that to count as “going for it.” This also prevents the “going for it” option from getting a boost as that decision lends itself to getting extra downs
The Results
In the regular season the play finder came up with a pretty darn good list of plays that meet all of the criteria above. Luckily, there’s a list of toggles for play type that lets us choose between kneel downs and all other types of plays.
Going for it
The play finder gives us a list of 145 plays from 2009 to present day that meet our criteria and the team elected to try to move the ball. So that’s about 14 times a year that a team has a touchdown or three advantage right before the half and they elect to try for another quick score before the break.
Take a knee
Let’s flip the toggle and see how often teams make a decision similar to Sean McDermott.
Three times. Three times a coach in our defined parameters of similar plays elected to take a knee. One of them is Josh Allen doing it last Sunday.
One problem
There’s one problem that we need to address—it’s not tweaking the parameters. I ran it a bunch of ways and landed on the criteria above for a variety of reasons. Rather than bore everyone, ask me in the comments if you want a rationale on any of them.
The biggest issue is that we want to know how many of these plays were the first play of the drive. Unfortunately the play finder does not allow that information to be easily sorted. On the plus side, though, the sample size for kneel downs makes this easy to figure out manually. What I did was review all three kneel downs to find out how many were the first play of the drive. Then I hunted through the “go for it” plays to compare, crossing my fingers I didn’t need to go through all 145.
The results
Look, it’s not even close. All three kneel downs were the first play of their respective drives. There’s no way I was going through all 145 other plays so I decided I’d see how many plays in the list I needed to double the total and that would be an acceptable extrapolation.
I found my goal of six instances in the first nine plays listed. The sample size is a very imperfect extrapolation but I have a high degree of confidence that there’s really no contest between the two options. And if we’re being completely fair in our strategy, I wouldn’t include the example from this Sunday. Since we’re comparing McDermott to all the other times a similar situation presented itself, it’s skewing the results in favor of taking the knee by including our comparison play in the sample.
I’m open to questions on the method, but as of right now I believe this is a fair look and the results are lopsided. When faced with a similar situation, the vast majority of teams will try to put more points on the board.