Kirk Cousins and the History of Mega-Comebacks

I dare you to say that again...

He is also a horrible punt returner. 

Carolina ended up winning 31-24 and never saw their WP drop below 90%.

Eh, "simple WP" is heavily biased in those situations, because you've always got "extreme high leverage yet unlikely event" in there. So it's not "a string of 60% plays" it's "a string of 90+% plays plus one 10% play" or something. It's a tail probability issue - teams that are capable of converting low percentage plays frequently don't get into that situation often, so trying to judge "how likely is it that Dalvin Cook has a 60+ yard TD?" is really, really hard.

I think it's a bit tough to guess what the Colts intentions were in this game specifically because, well, the Colts offense sucked the whole game. This isn't a comeback no one could see coming: Minnesota had a ton of high-leverage, random single plays in the first half. By EPA that first half would've looked terrible for the Vikings, but overall it was just the equivalent of a few "run for no gain" plays which isn't that bad. They just happened to occur at extremely bad times.

DVOA measures efficiency. So yes, if a coaching staff’s goal is to be inefficient, then DVOA stops being an effective measurement of success. That’s not really a knock on DVOA, though; generally speaking, being inefficient at your job is a bad thing, and it’s not as if eating clock and being efficient (on a per-play basis) are somehow mutually exclusive. 
 

Also, I believe DVOA does give an adjustment for blowouts. E.g. giving up 10 yards on 3rd and 8 is bad, but it’s comparatively not as bad when you’re up three scores vs being tied. Probably what DVOA is seeing is that the game mathematically stopped being a blowout after a certain point, whereas human observers tend to project a narrative and might not make the same distinction. 

OK, this is a fun exercise for a slow work day before Christmas...

Let's assume that each kneel play takes 3 1/3 seconds, loses one yard, and the Colts run the full 40 seconds between plays, to make the math a little simpler; that means each "drive" loses three yards and uses 2:10 of the clock between the start of the drive and the fourth-down snap.

The Colts first got the ball on their 34 at 13:57 in 3Q.  Three kneeldowns would have given 4-13 from their 31 at 11:47, instead of 4-7 from their 48 at 11:23.  The resulting 40 yard punt and fair catch then would have given MIN the ball on their 29 at 11:40 instead of their 12 at 11:16.  (And here we can see the start of the butterfly effect, because a punt from your own 31 is likely to be longer (and lower so more likely to be returned) than one from your own 48 - but let's ignore that for the moment.)

The Vikings then spent 2:54 to score a TD.  And let's assume that they would have spent the same amount of time to drive 71 yards instead of 88, so when Flowers returns the ensuing kickoff to the 40, the Colts' next drive starts 24 seconds earlier (8:46 instead of 8:22).

Instead of taking 3:23 to drive to the MIN 34 and kick a FG at in the next drive, they use 2:10 to take three knees and punt from their 37 with 6:36 on the clock.  Let's assume a 42 yard net punt (game average was 41.7) using 10 seconds, so MIN gets the ball at their 21 at 6:26 (instead of at their 25 at 4:53 following the kickoff).

That's close enough to the actual next drive that we'll plug it in - MIN uses 3:34 to drive down the field and score another touchdown, making the game 33-14 (since IND punted instead of kicking a FG).  The kickoff is run back to the IND 33 using another 6 seconds, so the next Colts drive starts at their 33 at 2:46 in 3Q (instead of 1:13).

Three kneeldowns result in 4-13 at their 30 with :36 in 3Q (vs 4-7 at their 36 with :15) - close enough that we'll use the same Haack 38 yard punt and 13 yard return, starting MIN's next drive at their own 45 with :28 on the third quarter clock.

Again we'll assume the Vikings' next TD drive (55 yards instead of 61 yards) would have taken the same 2:08, and the kickoff would have been returned to the IND 17 using another 6 seconds, so the next IND drive starts on their 17 21 seconds earlier (at 13:14 instead of 12:53).

Three kneeldowns give them 4-13 at their 14 with 11:04 left (instead of 4-6 at their 32 with 10:42).  Using the same punt, return, and MIN penalty, the Vikings start their next drive at their 31 with 10:48 left in the game.

They take 2:20 to drive to the IND 41 and throw a deep pass that's picked off at the IND 2, so the Colts start their next drive at their 2 with 8:28 (vs 8:02).  We can't use the same kneeldowns because the Colts aren't going to take a safety, so let's assume a combination of kneeldowns and QB sneaks that net 0 yards over 3 plays and take the same 2:10, and they're punting from the 2 with 6:18 left (instead of from the 4 with 7:13).  A 46 yard punt is fair-caught and takes 7 seconds, so now the Vikings are starting on the IND 48 with 6:11, and the kneeldown strategy is starting to pay off, because the Vikings have 55 seconds less game clock (but only a 33-28 lead)...

The Vikes use 1:36 to drive down and score their next TD, so after the kickoff and touchback the Colts are starting on their 25 with 4:35 left.  Three kneeldowns give them 4-13 on their 22 with 2:25 left - but critically they avoid the fumble that gives the Vikings the ball on the IND 39 at 3:25.  Instead, let's give another average punt/return (42 yards, ten seconds) and the Vikings start their next drive at their own 36 with 2:15 left.

Minnesota then runs 8 plays in 33 seconds and turns the ball over on downs after gaining 9 yards, so the Colts get the ball back at 1:42 on the MIN 45 (the plays were all either incomplete passes or OOB, so the two minute warning wouldn't have changed the timing).  Now MIN will use its timeouts, so the three IND kneeldowns will only run 10 seconds off the clock, and IND will face 4-13 from the MIN 48 at 1:32 up by 5, where in real life they faced 4-1 at the MIN 36 at 2:31 up by 8.

In real life they went for it, but in the kneel-and-punt world they would have punted.  Let's again assume a 40 yard punt fair-caught (like the first actual punt of the third quarter), so now MIN is at their 8 yard line with 1:26 left and no timeouts. 

This is where, in reality, Cousins tossed to Cook who ran for 64 yards, and this is actually the first point (other than the fourth down decision by the Colts at the last drive) where this thought experiment diverges enough from reality that we can't just assume the same thing happens next.  Which is remarkable in and of itself.

I suspect the same play (which is not an unreasonable call from your own 8) wouldn't have gone for a TD - Cook is more likely to get run down trying to go 92 yards than 64.  But even if he goes ~64 yards and gets tackled, the Vikings then have 1-10 at around the IND 28 with about a minute left (between the time of the play and the time needed to get down the field and spike the ball, assuming he doesn't go out of bounds).

So it's not totally clear that the Colts would have won the game simply by taking a knee on every offensive play in the second half - but it's also not clear that they would have lost it, and it's amazing to me how similar the situations and results would have been...

I've been told the home-field advantage is now down to 2 points. So a 4- rather than 6-points swing.

It seems the Scott Mitchell Experience should serve as its own lesson.

\In true Lions fashion, Goff is actually Erik Kramer, and the Lions will acquire their Scott Mitchell.