Late-Game Leverage
NFL Week 14 - In the late stages of a poker tournament, as the number of big blinds starts to dwindle, skilled poker players recognize the necessity of taking binary action. ‘Push or fold’ becomes the optimal strategy. Similarly, fourth-down strategies in the late stages of a football game can become justifiably more aggressive. However, these decisions still require careful calculation and are often quite complex, especially when the game is close.
As the Buffalo Bills were mounting a comeback Sunday afternoon against the Tampa Bay Buccaneers, some difficult decisions arose for Sean McDermott during the fourth quarter and in overtime. Before we get to that, let’s look at some earlier decisions that were somewhat costly to the Bills in terms of their Game-Winning Chance (GWC).
| Clock | Score | Situation | Bills’ Choice | Optimal | Cost |
| 9:32 Q2 | Trailing 10-0 | Fourth-and-goal, Bucs’ 3-yard line | Field goal attempt | Go | -2.2% GWC |
| 12:35 Q3 | Trailing 24-3 | Fourth-and-2, Bills’ 45-yard line | Fake punt, Go | Go | +0.6% GWC* |
| 2:19 Q3 | Trailing 24-10 | Fourth-and-3, Bills’ 45-yard line | Punt | Go | -1.4% GWC |
*While we agreed with going for it, the fake punt strategy seemed unnecessary. It was a mathematically defensible decision to attempt the first down, and simply putting the ball in Josh Allen’s hands seemed more appropriate.
Now moving on to the Bills’ final drive in regulation and the first possession of overtime.
| Clock | Score | Situation | Bills’ Choice | Optimal | Cost |
| 1:48 Q4 | Trailing 27-24 | Fourth-and-4, Bills’ 40-yard line | Go | Go | +19.7% GWC |
| 0:22 Q4 | Trailing 27-24 | Fourth-and-2, Bucs’ 7-yard line | Field Goal | Field Goal | +7.9% GWC* |
| 8:58 OT | Tied 27-27 | Fourth-and-4, Bills’ 31-yard line | Punt | Punt | +2.1% GWC |
With no timeouts remaining, the Bills’ decision to go for it on fourth-and-4 from their own 40-yard line is a no-brainer. Anything else is suicide. The final choice in regulation to kick the field goal is far more interesting. However, there just wasn’t enough time on the clock to justify going for it, as the Bills were out of timeouts. Two timeouts would have slightly tipped the decision in favor of going for it. Finally, the fourth-and-4 choice in overtime is about as highly leveraged as a decision can be and surprisingly close. The Bills correctly chose to punt, which produced 2.1% more wins on average, per the simulation. If it was a fourth-and-2 rather than a fourth-and-4, it would have been a break-even decision to go for it.
Comments
11 comments, Last at 16 Dec 2021, 7:47pm
#1 by Seattle-Brian // Dec 14, 2021 - 12:51pm
Specifically with the punt decision at 8:58 OT Tied 27-27. Seems like a punt decreases the possibility of a loss much more than it increases the chance to win. Does a tie count as a win to GWC? Or is there also a GLC (Game Losing Chance) stat that would be interesting on this one?
#2 by JacqueShellacque // Dec 14, 2021 - 2:37pm
The final choice in regulation to kick the field goal is far more interesting. However, there just wasn’t enough time on the clock to justify going for it, as the Bills were out of timeouts. Two timeouts would have slightly tipped the decision in favor of going for it.
How does this square with the universally derided (by analytics types) choice of kicking an extra point after scoring a TD when down 14, because it's 'only' playing to get to OT?
Also there's simply no way this can be quantified, as the only comparable situation is probably the Miracle at the Meadowlands or some similar once-a-century play. Not a great basis for calculation, nevermind as precisely as '7.9%'. This is a possible game-ending scenario and is therefore discontinuous, use of probability distributions here is inappropriate.
#3 by Pat // Dec 15, 2021 - 9:15am
How does this square with the universally derided (by analytics types) choice of kicking an extra point after scoring a TD when down 14, because it's 'only' playing to get to OT?
Because the chance for just winning in regulation here is very low if they just get the first down, which means they're just going for the TD directly, and that chance is low.
The "kick XP down 14 after scoring a TD" is a bad prima facie decision because the expected value of going for 2 twice is still 2 points, so going for it doesn't really sacrifice anything - the most likely result is still overtime. (You can argue that there's some value to keeping your opponent passive, as they might be more aggressive only up 6, but that's a 'playing your opponent' thing, like a bluff in poker).
Also there's simply no way this can be quantified, as the only comparable situation is probably the Miracle at the Meadowlands or some similar once-a-century play.
Are you misunderstanding the situation here? If the Bills had gone for it and succeeded, they didn't need a miracle play to win. They just would've had to score a TD. That's happened a ton of times in NFL history. If they had gone for it and failed they would've needed a miracle play, but you can just set that chance to 0 and there's still a comparable situation.
The chance of the Bills scoring that TD can easily be estimated by just looking at cases where the decision's forced (down 4 instead of 3), since defenses will be basically playing things exactly the same.
#4 by jheidelberg // Dec 15, 2021 - 10:54am
A whole article was written about the two point conversion after scoring a TD down 14 and I refused to comment. This discussion is so silly, it is a proven mathematical fact that you should go for two. I have saved my comment for now.
Zac Taylor, who already had an awful game making decisions on 4th down as written about by EDJ sports, had the classic down 14 then score a TD decision in the 4th quarter. Being Zac Taylor he went for an extra point, his team did indeed get another TD and extra point in regulation to send the game to OT and his team lost in OT. He blew the added and proven mathematically opportunity to win in regulation by not going for two after the first TD. Not to mention that his team could have lost in regulation by missing the second extra point.
#5 by Pat // Dec 15, 2021 - 11:12am
This discussion is so silly, it is a proven mathematical fact that you should go for two.
It's really a small improvement in game-winning percentage. In terms of relative improvement, it's huge (a 50% improvement) but because your chances are already pretty small at that point, in absolute terms it's not a huge deal. Saying "OMG that was a terrible decision" to go for it is a little overblown - it absolutely pales in comparison to choosing the right play on the next offensive drive or calling the right defenses for the next plays.
So in some sense, if you're a head coach, you score the first TD and you're grabbing the Surface tablet scanning their offense trying to figure out how you're going to stop it, if an assistant runs up to you and says "coach! coach! we need a 2 point play to improve our chances by 3%!" and you say "shut up, just kick it, I'm busy" that's totally the right decision to make.
(All of that ignores the fact that the correct decision in that case would be to be a good enough organization to delegate that stuff, but hey, in that case you are talking about the Bengals)
#7 by jheidelberg // Dec 15, 2021 - 1:44pm
Since you can not win without the second TD, it must be assumed in making the decision. So we go back to the basic equation:
Assuming 48% to make a two and 95% to kick an extra point:
Go for two first and succeed: .48*.95= 45.6% win
Go for two first and fail: .48*.48= 23.0% lose
OT 1-.456-.23= 31.4%
This is way more than 3% but I understand you are counting the fact that it may not come down to this anyway.
The decision has no relevance at all if the play does not go accordingly. That is the whole issue with coaching decisions, Houston rarely makes the EDJ top 5 or bottom 5 because they do not play many close games. If I am the owner, I still expect proper coaching decisions whether or not I have a good team or not.
Bottom line: Coaches are paid millions of dollars, the difference between Harbaugh and Taylor's coaching decisions is monumental (despite Taylor having his best year and not being at the bottom of the coaching decision ranks).
Harbaugh's value added on coaching decisions justify his entire salary. If I pay my coaches millions I want that extra 3% every time, Harbaugh is worth every penny that he is paid on coaching decisions alone, not including all the value that he adds by hiring a great staff to have the record he has during his tenure when Joe Flacco has been his QB the majority of the games that he has coached, game planning, etc.
I TOTALLY AGREE WITH YOU THAT THIS STUFF SHOULD BE DELEGATED. For example, Tomlin is a great coach, but could sure use some help stopping his punter and FG kicker from becoming regular players and 4th and 2 and 4th and 3.
#8 by Pat // Dec 15, 2021 - 4:07pm
Since you can not win without the second TD, it must be assumed in making the decision
Assuming the TD means you're calculating relative win chances. Like I said - the relative improvement is huge. The absolute improvement is tiny, since the chance of "stop + score" is somewhere between 15-20%. So multiply all of those by 15-20%, and the difference between them is a few percent, like I said.
Calculating relative win chances assumes that "stop + score" chance is the same in both "go for 2" and "kick it" situations, and that's a very big assumption. You're welcome to think it's obvious and must be true. It's still an assumption.
If I am the owner, I still expect proper coaching decisions whether or not I have a good team or not.
Proper coaching decisions involves maximizing your absolute chance, not your relative chance. You don't have infinite time to choose every decision. Think of it like speed chess: it doesn't necessarily matter if you make the absolute best choice.
I TOTALLY AGREE WITH YOU THAT THIS STUFF SHOULD BE DELEGATED.
The issue with delegating is that the person you're delegating to has to be at least similar in ability to you, and that's not always possible. Belichick, Harbaugh, and Reid have had decades to sculpt their staff the way they like it and people stay there well past when they should likely leave for better jobs. That's just the benefit of long-term continuity. Taylor might not have that kind of option. Again, it's the Bengals.
#10 by jheidelberg // Dec 15, 2021 - 5:14pm
Proper coaching decisions involves maximizing your absolute chance, not your relative chance. You don't have infinite time to choose every decision. Think of it like speed chess: it doesn't necessarily matter if you make the absolute best choice.
When there are only two choices that can be made, A or B it is reasonable to assume that a coach being paid millions of dollars would make the correct decision.
Coaches should be practicing just like players. It is getting late in the season. Every single coach in the playoff hunt should know pre-game whether or not a tie is an improvement or a detrimental to his teams playoff chances. Thus, when his team scores a TD with a few seconds left to be down 1 point, he should know whether or not he should go for two or play for OT, where the chances of getting a tie or win are greater than getting the win in OT.
I TOTALLY AGREE WITH YOU THAT THIS STUFF SHOULD BE DELEGATED
Congratulations Pat, you even find a way to argue with someone when they agree with you. I try to diffuse a situation in which we have a disagreement. That strategy is a clear failure. You are absolutely incorrigible.
#9 by nat // Dec 15, 2021 - 4:18pm
This is a possible game-ending scenario and is therefore discontinuous, use of probability distributions here is inappropriate.
It does not matter when the game ends. It only matters if you are winning or losing when it does end.
The resistance to the “go for 2 early” advice has always befuddled me, because understanding it is decidedly not deep analytics. If you get the 2 points early, you’ve set yourself up for a winning TD, which is twice as good as setting yourself up to get to OT. If you fail to get the 2 points, you still get another chance to get to OT after a TD drive. That combination is a lot more valuable than only trying for OT.
As for the situation in this game, it is not at all like going for 2 early. It’s a lot more like going for 2 and the win on the last play of regulation. Except that from the 7 yard line and with insufficient time to do more than kick if you convert but don’t score, it’s a worse deal.
#6 by TyWill40 // Dec 15, 2021 - 11:50am
I love this column. Could you do me a favor? There was a fourth down decision by in the 2003 NFC Divisional Game between the Packers and Eagles that I consider the worst in NFL history. Could you calculate the Win Probabilties on that play?
The play occured at the 2:30 mark of the fourth quarter. The Packers led the Eagles by 3 points. The Packers had just held the ball for over five minutes with a successful "run only" drive that began on their own 16 yard line. They now faced a fourth and less than one yard to go at the Eagles 41 yard line. The Eagles had only one timeout left.
Instead of going for it and essentially clinching the game, Mike Sherman inexplicably decided to punt. The punt went into the end zone, so the Packers gained only 20 yards with the punt. To add insult to stupidity, on the Eagles first play they gained 21 yards thus wiping out any advantage the punt gave the Packers in one solitary play. The Eagles went on to win the game in overtime.