Crazy thought of day: Cowboys should have missed extra point on purpose with 2:32 left so they'd be up 3, not 4

OK. You are missing something Rojo.

If a team puts themselves up by 4 points, then the other team has no option, but to go for the touchdown on 4th down (let's call a succesful touchdown event A).

If a team puts themselves up by 3 points, then the other team has the option of either kicking the field goal to tie (lets call this event B) OR going for the touchdown for the win (again, this is event A).

If a team puts themselves up by 4 points, their chances of winning = Prob (Event A not happening) = 1 - P (A)

If a team puts themselves up by 3 points, their chances of winning are the smaller value of "1 - P(A)" or "1 - .5xP(B)". That .5 is in there because of overtime of course. ANYWAY, if a team puts themselves up by 3 points, this is worse SIMPLY because they are giving the other team the OPTION of picking the best choice from 2 different options. Of course, you have to assume the other team will pick the best option. If the team is up by 4, they don't give the other team the option to pick the best of the 2 choices. That's why it's the smaller of those 2 values, because you're letting the other team pick.

I didn't read any of the posts except the OP, but Rojo, you clearly made a miscalculation somewhere. If I looked at exactly what you did, I could tell you, but I didn't look in detail, and you should be able to figure it out yourself anyway. IF you actually didn't make a miscalculation, then basically what you're saying is the OTHER team, when down by 3, is actually making the wrong choice of the two picks, and should always be going for the touchdown when down by 3. For some reason, I doubt that is the case (although I guess it could actually be possible).

Hello football fans. A few updates. Brian Burke of http://www.advancedfootballanalytics.com/ got back to me and it looks like my comment about Pro Football Reference's model having a different result can be ignored:

The question then is really how valid are Burke's advanced nfl stats numbers. I really think a key part of it was exactly how much time was left - 2:32. If there was 1:30 left, I'd much rather be up 4 than 3, but 2:32 left and I think it's possible my crazy thought is actually the correct one - it's better to be up 3 than 4.

That being said, Burke says he's got a model he uses for clients that isn't as certain that being up 3 is better than being up 4. Here's what he wrote:

So it's probably the right choice but we may not see it used any time soon.

rojo wrote:

There is a scenario where the head coach would look good. Detroit drives down the field gets to about the 5 yard line. Faces 4th and 1 and decides to kick the field goal to go to OT. The Cowboys win in OT.

I don't think so. If that scenario played out, everyone would be saying that the Cowboys won despite Jason Garrett's outlandish decision.

By the way rojo - even though I think the premise of your thread is nuts (you even acknowledge it as crazy) I applaud you for starting an interesting discussion. My suggestion - why don't you send your idea to Peter King at CNNSI? He eats up this kind of stuff.

Bad Wigins wrote:

Conundrum wrote:

I would suggest that before we try to reason out the cause for a particular statistical pattern, we should determine if the pattern is consistent with reality.

First we should identify those patterns instead of just spitting out numbers allegedly based on them.

I've already shown those numbers are inconsistent with published points-per-drive and scoring % statistics.

True. We need to determine if there actually is a pattern.

I do give props to Rojo for starting an interesting thread and going out on a limb on a subject that he knew he would get some blow back on. He was a good sport in letting people take have some fun with the ideas of the owner of this site.

rojo wrote:

Burke says he's got a model he uses for clients that isn't as certain that being up 3 is better than being up 4. Here's what he wrote:

"My newer model I use for clients isn't as confident about that phenomenon, so I'm hesitant to go out on a limb with it. "

So it's probably the right choice but we may not see it used any time soon.

That's a very strange conclusion.

A) Basically, with the earlier disclaimer you said he had, and now this new model saying it is uncertain if it's better to be up 3 or 4, he is NOT saying: it's better to be up 3 than 4. He's not sure.

B) as malmo and others have stated, there is no reason to think this guy's prediction calculator is perfect.

"so probably the right choice" is not correct. "Probably similar odds of winning when up 3 or 4 at that point" is likely a better conclusion, and in that case, no reason to miss the extra point.

Joe the Jeweler wrote:

If a team puts themselves up by 4 points, their chances of winning = Prob (Event A not happening) = 1 - P (A)

If a team puts themselves up by 3 points, their chances of winning are the smaller value of "1 - P(A)" or "1 - .5xP(B)".

We are talking about human beings not computers. The whole premise of Robert's point is coaches operate differently when down 3 than down 4.

So 1 - P(A) is no the same when down 3 as compared to down 4. When down 4 the team is going to go for it on 4th down and not settle for the field goal. They are more likely to score a touchdown (and win the game) than when down 3. I think everyone agrees to that point.

So odds of touchdown to win game are greater when down 4 than down 3.

This lower odds of scoring a touchdown when down 3 than when down 4 are offset by the chance that the team can kick a field goal and then have a 50% chance of winning in overtime. In an ideal world, the coach would be able to balance these odds perfectly so being down 3 is better than being down 4. What Rojo is arguing is that historically coaches have not done this very well and are too conservative and more likely to lose when down 3 than down 4. They think "I only need to get to the 30" and I can kick a long field goal. They therefore might be too conservative in the play calling. Once they get to the 30 they may stop taking shots down the field (Jason Garrett used to do this and act like a 50 yarder was automatic). A coach rarely is going to get fired for trotting out his guy for a 50 yarder but if he takes one more shot down the field and it gets picked off he might get criticized.

Let's say down 4, a team scores a touchdown 30% of the time to win.

Down 3 they score a touchdown 15% of the time to win and 30% of the time they kick a field goal. The odds of winning the game when down 3 would still be the same (15% + .5(30%) =30%,. I'm open to the possibility coaches get this calculation wrong. It has been shown coached definitely go for it too little on 4th down to begin with so I don't see why everyone is so opposed to the idea that coaches might be too conservative in their play calling when down 3. Then things can get more complicated with field goals leaving time on the clock for the other team to come kick a field goal, etc.

Noted Letsrun and Deadspin poster Raycism has made this debate front page material on Deadspin.

http://sidespin.kinja.com/x-the-point-the-danger-of-the-4-point-fourth-quarter-l-1678062049/+samer

Dragon Runner wrote:

By the way rojo - even though I think the premise of your thread is nuts (you even acknowledge it as crazy) I applaud you for starting an interesting discussion. My suggestion - why don't you send your idea to Peter King at CNNSI? He eats up this kind of stuff.

Thet sounsd like two much werk. i just want too wach me sum tV.

Wejo,

I don't think it's an honest approach to act like the Cowboys are the only team with the computer, which seems to be a key premise of the original post and the computer model you presented. The model assumes one team can accurately forecast inefficiencies in how the other team will play. But if Jason Garrett can run a fairly accurate model in your hypothetical situation, then why wouldn't Jim Caldwell do the same and simply treat a 3 point deficit as 4 points until a situation dictated that kicking the field goal became the favorable option? (Granted, it's hilarious to picture Caldwell trying to use a rotary phone, let alone a computer.) In the event that one assumes the opposing team can also model and mostly optimize their choices, any accurate model would logically favor a 4 point lead for the Cowboys as described in detail earlier due to the Lions having more possible winning outcomes from a 3 point deficit.

rojo wrote:

I really think a key part of it was exactly how much time was left - 2:32. If there was 1:30 left, I'd much rather be up 4 than 3, but 2:32 left and I think it's possible my crazy thought is actually the correct one - it's better to be up 3 than 4.

It's even worse than usual with that much time left, because it allows them to march up the field with short gains into FG range and then still have time to take shots into the end zone. It's harder to defend because you can't focus on long passes or they will nickel and dime their way to FG range easily. And they will run down any remaining time before kicking anyhow.

b runs wrote:

Noted Letsrun and Deadspin poster Raycism has made this debate front page material on Deadspin.

http://sidespin.kinja.com/x-the-point-the-danger-of-the-4-point-fourth-quarter-l-1678062049/+samer

That guy's the best.

Y'all (at least most of you) are missing one very important fact:

Variance is a BITCH!

Actually, in this instance, you go for the two point conversion so you can get a 5 point lead. When the other team scores a TD + extra point, all you need is a field goal to regain the lead!

rojo wrote:

What people seem to be forgetting is up 3, the defense still gets a chance to make a stop. In fact, the Cowbos defense did just that. It stopped Detroit before Detroit got in field goal range.

No.

The Lions were in Prater's range when they were stopped on 4th down at the end.

Had they been down by 3 they were in position to tie it.

The 4 point lead forced the Lions to go for it and they failed to convert, giving the win to the Cowboys.

Rojo's idea has made it to the big time now! On my drive home from work last night, Steve Czaban was talking about the story featured on Sidespin. He sounded like he was in complete agreement with Rojo!

Ok. Even if we accept that being up by 3 has a slightly better EV than being up 4 - which you clearly have not even come close to proving in any way at all - the idea is still ludicrous. Here's why:

Probability and expected value only come into play over the very, very long term. There are poker theorists who now believe that 1 million hands is not enough to rule out losing due to variance alone.

An NFL team can expect to find themselves in such a situation maybe 1 or 2 times a year. The team is not interested in winning as many of these scenarios out of 1 million as possible because it will never be there a million times. The team is interested in this one scenario. And with a sample size of 1 your whole premise is idiotic, as it is clearly easier to stop a team from scoring a touchdown than from kicking a field goal. Basic redzone stats will show this.

Your whole post reads like those shysters trying to sell idiotic football bettors on why they can beat the spread, with crap like "The cowboys are 58 and 2 since 1971 on the road on Sundays when Methodists have church suppers and the moon is in the first house of Uranus."

Star wrote:

No.

The Lions were in Prater's range when they were stopped on 4th down at the end.

Had they been down by 3 they were in position to tie it.

The 4 point lead forced the Lions to go for it and they failed to convert, giving the win to the Cowboys.

Your argument supports Robert's point.

It was 4th and 3 at the Cowboys 42. That's a 59 yard field goal. He makes that what 10% of the time? So lets say they attempt the field goal attempt there.

10% of the time they get to go to overtime where they have a 50% chance of winning the game. So 5% of the time they win the game if they attempt the field goal. 95% of the time they don't. The math would say definitely go for it there on 4th and 3.

What if they are 5 yards closer? A lot of coaches are tempted to put it on the kicker instead of being known as the guy who went for it when they could have kicked a 50 yarder (which if made would only give them a 50% chance of winning)

No-No-It's not wrote:

Ok. Even if we accept that being up by 3 has a slightly better EV than being up 4 - which you clearly have not even come close to proving in any way at all - the idea is still ludicrous. Here's why:

Probability and expected value only come into play over the very, very long term. There are poker theorists who now believe that 1 million hands is not enough to rule out losing due to variance alone.

Once again, if you're willing to accept the premise that the expected probability of rojo's strategy is better, and the cost of losing and benefit of winning by both strategies is equal, then your statement is wrong as to which is the best strategy.

[quote]No-No-It's not wrote:

An NFL team can expect to find themselves in such a situation maybe 1 or 2 times a year. The team is not interested in winning as many of these scenarios out of 1 million as possible because it will never be there a million times. The team is interested in this one scenario. And with a sample size of 1 your whole premise is idiotic, as it is clearly easier to stop a team from scoring a touchdown than from kicking a field goal. Basic redzone stats will show this.

[\quote]

You just appealed to the same type of argument that you just claimed was wrong "Basic redzone stats" are still stats

[quote]No-No-It's not wrote:

Your whole post reads like those shysters trying to sell idiotic football bettors on why they can beat the spread, with crap like "The cowboys are 58 and 2 since 1971 on the road on Sundays when Methodists have church suppers and the moon is in the first house of Uranus."

[\quote]

You just changed your argument to _not_ accepting the premise. I think we're (including brojos) all in agreement that the premise is likely wrong.