It's not the same. You expect that the other team will score with so many possessions remaining.
And when they do, they will be able to capitalize on the information gained from your 2 point attempt.
In terms of expected value, 2 point conversion attempts are roughly par with extra points, though with so few scoring events in the game you would expect large variance.
So outside of scenarios where it is pure upside (1 point doesn't matter without many more scoring events) teams will opt for the lower variance outcome. This matches practice pretty well IME.
> It's not the same. You expect that the other team will score with so many possessions remaining.
If Baltimore special teams doesn't let a punt return TD happen, Houston had one FG to their name... the entire game.
I don't expect a team will always score, absolutely not a team playing against one of the best defenses in football history. So, I respectfully disagree with your thesis. :)
You may not personally expect it, sitting on the sidelines. But the people on the field do have to expect it. They can't make sound strategic decisions while operating under an assumption that the opposing team will never score a touchdown.
The issue is that a coach going for 2 in the first quarter and lose because of it will look very bad and the media will always look at those and ask why you didn’t go for easy 1. Coaches aren’t gonna put their job on the line to look cute
Well it could very well be a slight inefficiency as a result of tradition or inertia or whatever.
I also suspect that the probability of converting for 2 goes down with each subsequent attempt. You can expect teams to have a few plays installed specifically for critical 2PC attempts, and you may not want to burn those plays for marginal gain early in the game.
these are the assumptions they are attempting to debunk:
- All 2-point conversions are equally likely to succeed.
I have seen no evidence that all 2-point conversions *aren't* equally likely to succeed, and even if they were two bites at that apple gives you better odds than just one
- You will stop the other team from scoring.
regardless of whether you choose to go for it or kick the XP, you still have to stop your opponent on their subsequent drive. if they do so much as kick a FG it's game over already, so it doesn't matter.
- You will get a touchdown on the subsequent drive.
you have to score a TD or the debate doesn't matter so it's irrelevant to this argument
- The clock expires after that.
again we have to assume we are preventing the opponent from scoring any additional points because the debate makes no sense without that caveat
- There are even odds of winning in overtime.
getting to OT in this scenario is the fallback, while it's the best outcome of just kicking the XP twice in a row
Surely implicit in the statement is "all 2-point conversions in the 4th quarter by the same team" since we are comparing 3 possible 2-point conversions in the 4th quarter by the same team. (2pt conversion when down by 8, 2pt conversion when down by 2, 2pt conversion when down by 1).
Doesn't this article basically show the exact opposite of what it's claiming? It cites simulation data showing that going for 2 here only increases the chances of winning by an absolute 1.7% and uses that to claim that that's a small increase, but it's actually not. The winning percentage of kicking for an extra point is already at 11.2%, increasing that by 1.7 is quite large (and is actually a larger increase than the relative 12% increase the article says is ludicrous).
If you're down 8, with limited time remaining, you should always go for 2 to tie the game, instead of still being down by 1. You'd still need a field goal to take the lead, and that would still be the case if you miss the 2pt conversion.
the context is "yes, you were down by 14, but your team just lined up and scored a TD on the last play. you have to make the decision whether to kick the XP or go for 2. since you just scored 6, you are now down 8"
it is confusing because the conversation is taking place after you just scored
I don't really understand the disagreement. The main disagreement seems to be that others say it increases the win probability by 12% assuming you score the needed touchdown afterwards, and the author says that it's not really 12% because there's a 90% chance you won't score the touchdown? On its face, I tend to agree with the others that that's irrelevant.
In the linked blog post with the actual statistics in it, she argues that it only increases your chance from 11.2% to 12.9%, so it only really increases your chances by 1.7%. But that just seems to be dishonestly reframing the numbers so people think a 1.7% absolute increase is nothing. But that really is a 15% increase in win chance. It sounds like the data says the 2 points is even better than its proponents claim?
His only other argument is that failing the conversion psychs a team out, which I respect, but that's still just a bare assertion. I don't know why you'd get so mad at statistics for ignoring your feelings. People are complex and absent data otherwise, I would think it just as plausible that the added pressure and imminentness increases performance.
No, that's dishonest. Probabilities are already ratios between magnitudes (desirable vs. undesirable outcomes), so when comparing them you should be subtracting them, not dividing them. A relative increase in probability of 15% doesn't tell you anything if you don't already know the base probability. Also, why go for the 15%? You could also say that the loss probability is reduced by 2%. But I guess you don't get a nice juicy number that way.
If you are given the option of taking one lottery ticket or two lottery tickets, is it dishonest to say that taking two lottery tickets would increase your odds of winning by 100% over taking just one? Do you need to know the odds of a single ticket winning to make that claim?
I think you're using a restrictive and non-standard interpretation of percentages. They can be used to convey overall probability, but also a change in probability. Only allowing one usage is just a communication deficit.
> If you are given the option of taking one lottery ticket or two lottery tickets, is it dishonest to say that taking two lottery tickets would increase your odds of winning by 100% over taking just one?
Not in that particular case, because no one would believe in an increase of 100% (although it would still be preferable to say that it doubles your chances of winning). But if you're at a slot machine and you can put in an extra dollar for a 10% extra chance to win the jackpot, and in fact you chances increase from 0.001% to 0.0011%, would you think that's honest or scummy?
>They can be used to convey overall probability, but also a change in probability.
They can be. They shouldn't be used this way because it lends itself to miscommunication.
> But if you're at a slot machine and you can put in an extra dollar for a 10% extra chance to win the jackpot, and in fact you chances increase from 0.001% to 0.0011%, would you think that's honest or scummy
Casinos in general are scummy, but I can't fault with the phrasing. What's the threshold at which the world is allowed to use percentages in the way they're meant to be used, without having to be accused of being dishonest?
IMO they should never be used this way period, or at least not without also indicating the values with and without the change. You'll be accused of being dishonest if you use a number that is convenient to your argument (e.g. "buying this is good value!") but which a significant number of people will interpret incorrectly.
You're right that in situations where you're still rounding to zero, talking about a 100% increased chance of winning is still basically zero.
It's important to distinguish the scummy behavior from the actual statistical practices, as data literacy helps people see through numerical slight of hand.
> They can be. They shouldn't be used this way because it lends itself to miscommunication.
I think you are backwards from how other people regularly faced with this ambiguity have decided to communicate, given that e.g. finance created a separate term to talk about changes in percentages.
Using that terminology, the win chance has gone up by 160 basis points, which is an increase of 15% from its previous value.
I don't see how that's in contradiction with what I'm saying. Expressing a change in a percentage in a while new unit or terminology seems ideal, because someone who's never seen it before will not mistake it for some other thing they already know. It doesn't lend itself to confusion at all, unlike expressing the change also in percentage.
Since you're talking about scummy behavior and miscommunication, I'd like to point out one place that TFA does so:
> Under completely unrealistic assumptions, G42D8 improves the chances of winning by less 1%. If there is any uncertainty in these calculations at all, then you might even be putting yourself at a disadvantage!
This is implying (without out saying) that a tiny error in e.g. the probabilities of converting a 2 point conversion will make things a disadvantage. After all, what if there's only a 46% (instead of 48% from Ms. Albert's article) of making the 2-point conversion, that must surely lower your win chances by more than 1 percentage point, right?
The answer is "of course not" becuase that just lowers the win probability by a tiny absolute amount as well. In fact this can be seen in the flow-chart that TFA hates so much. The author is either being scummy and disingenuous, or has confused himself by mixing relative and absolute percentage changes.
I can't say what the intention was. It certainly seems odd to attempt to play up the disadvantage when it seems clear it's basically inconsequential. It kind of reads like he's against it for seemingly irrational reasons.
Win probabilities are not ratios between magnitudes of desirable and undesirable outcomes. They are the fraction of total possible outcomes that are desirable. Otherwise a 1/3 win probability would be 1/2 and a 9/10 win probability would be 9.
One number never fully captures the reality. If I have a net worth of 1 penny and I earn a second penny, it is 100% true to say I've doubled my net worth. It also doesn't tell the full story.
If a coach can take an action that would move the win probability from 0.01% to 1%, it would be malpractice to not do so, and correct to say they've increased their win chances by a factor of 100. It's also correct to say they're probably going to lose.
I agree with you. She may be technically correct (the best kind!) but only in post hoc analysis of completed games. That should have 0 bearing on a coach's decision making during a game, since from the in-progress perspective it's still a 12.5% increase in win probability. So from a coach's POV, what's the worst that can happen: not getting the conversion and still losing by 1 (assuming :00 on the clock and no additional opponent possession). What's the best that can happen: tying and having an opportunity to play for the win in ensuing possessions.
Depends on the time left and the calibre of your qb and receivers if you have enough time to drive down to kicking range if you fail going for 2. The opposing team may have a great run game and will run out the clock. Usually with lots of time left you always go for 1 as that’s a high percentage shot and 2 is low
The context is "should I go for the 2-pt conversion?" When you're making that decision, you've already received 6 points for scoring the corresponding touchdown. The scenario presented is that you were down by 14, then you scored a touchdown to be down 8, and now have to decide whether to kick the PAT or go for 2 points. That's why it's correct to say "when you're down by 8". They're talking about a scenario where you'll still have to score another TD.
> ...the Bucs would eventually lose because Coach Todd Bowles forgot to use his final timeout.
Is this intended to be tongue-in-cheek? Reading such a statement (which logically implies that "if Todd Bowles had used his final timeout, the Bucs would have won") in a post about how other people's reasoning is insufficiently rigorous is off-putting.
No, it's not tongue-in-cheek but it's oversimplifying the situation. The Bucs might have been able to win if they had used their last timeout. It sounds like the coaching staff were morons and couldn't count to 3. However, they were given an extra timeout due to an unusual situation, and it seems that the coaches were either unaware of the extra timeout or forgot about it.
It's unclear. The Lions also weren't running down the play clock. Possibly everyone just accepted that the game was over and the Lions didn't run down the play clock and the Bucs didn't use their time out.
Or Bowles mentally marked the game as over and then didn't notice the Lions messed up with the play clock.
> So, the game clock showed 36 seconds remaining when Goff took a knee for the third and final time and the hugs and the handshakes, the celebrations and condolences, commenced.
> Now, had Bowles used his last timeout, maybe the Lions decide to punt instead of attempting a long field goal. But Bowles surmised a kick through the uprights would have ensued.
Basically, he would have had 36 seconds because of the clock mismanagement on the side of Detroit. Even with 0 timeouts, you can knock out 4-5 solid plays, which is a lot more than 0 solid plays. He goofed up.
Granting that this analysis is correct (it seems to be disputed), 36 seconds to play from deep in your own territory with an 8-point deficit is a far cry from a guaranteed win. If the Lions had kicked a successful field goal instead of punting, that would be an 11-point deficit.
Claiming that the Bucs lost because of this error is way off, they almost certainly would have lost anyway.
Yeah, it probably increases your absolute win probability by a percentage point or two. Just like the 2pt conversion strategy that TFA is ranting against
> Under these assumptions, Laura uses a computational technique called Dynamic Programming to evaluate all possible scenarios in the game tree. Once you factor in this more complex set of possibilities, the “62.5%” mirage disappears. Laura finds that if you are down by 2 TDs, G42D8 increases your probability of winning by less than 1%.
This is misreading the original source [1]. The <1% figure applies when you have not already gotten the first TD.
> We can examine the decision in more detail. When down by 8 with four possessions to go (which matches up with when Carolina went for a two point conversion), a team has one of two choices:
Down by 8 and facing a conversion opportunity means you were down by 14 and just scored a touchdown (6 points). It's more that 1 percentage point change (1.7%) not less (and as others have pointed out, it increases your win probability by about 15%), so there is still a misreading.
A psychological factor he doesn't mention: A lot of players, if they absolutely need to make a play to not lose the game, will dig deeper and play better. If the game is not on the line right now, they won't play as well.
In American football, getting the ball to the other side of the field is a touchdown and worth 6 points. Afterwards, you have the option of kicking the ball through the posts for 1 extra point (which is almost guaranteed) or trying to run the ball into the end zone in one attempt for 2 extra points. If you make a touchdown but you still have 8 fewer points than the other team, you can either take the 1 point and hope to score again then get 1 more point again to tie, or you can try for 2 points and hope to score again and then either take the 1 point for the win if you make the first 2 points or try the 2 points again then to tie it up. A "conversion" is scoring either form of extra points.
This is a very long article which seems to just reveal the author doesn't understand the difference between percentages and percentage points.
50% to 62.5% in the simplified calculation is a 25% improvement, not a 12.5% improvement. It's a 25% improvement on already low odds (after all, nobody is actually saying that it's 100% probability that your D will stop the next drive and then you will score another touchdown. It's just assuming that those things have to happen either way so the odds cancel out, and this is looking at win probability assuming that those things happen). Laura's calculations, for a similar scenario (4 possessions remaining), projects the overall probability of winning going from 11.2% -> 12.9%, which is a 15% improvement, roughly the same order of magnitude as the simplified example. I don't have the exact breakdown for a two-possession scenario, but in her slideshow she says this, which seems to be what the author is citing:
> If down by ...16 (before the 6 TD points), the strategy improves the win probability by <1%.
But I am convinced she is talking about absolute percentage points here (e.g. 10% to 10.9%), not a <1% change (e.g. 10% to 10.09%). Why? Because the slide has this before it:
> Rationale: if you are down by 15, you have to score two touchdowns while keeping your opponent from scoring
She is demonstrating that it's not a silver bullet given it's exceedingly unlikely you will win anyway.
Clock management is not just for the last minutes of the game, it is a fundamental part of the strategy for the entirety of play, starting before the game.
I enjoyed them from a not-really-even-a-fan perspective, somewhat like a Moneyball type of read.
What am I missing? It seems like this is based on assigning arbitrary probabilities to outcomes and calculating from there. I would assume that it was more useful to look at the outcomes of previous attempts by any team in the same situation and use that to guess how likely the outcome of this attempt is.
If you have to stop 'em once and score twice, you might as well try to win too? Kicking the extra points can only lead to a tie, but going for two affords at least a chance at a win in regulation.
You really do have an advantage by going for 2 first, and it's because you have gained information by doing so.
The possible choices you can make for 2 PATs are:
Kick, Kick
Kick, Attempt
Attempt, Attempt
Attempt, Kick
If you kick first, let's assume kicking is 100% successful, now your choices are:
Kick made => Attempt
Kick made => Kick again
These are both rational choices; attempting gives you a 50/50 shot to win, and kicking again sends you to OT with a 50/50 chance.
Now let's look at if you attempt first.
Attempt made => Kick
Attempt made => Attempt again (irrational!)
Attempt missed => Kick (irrational!)
Attempt missed => Attempt again
Some of these are irrational! If you made the conversion, just kick it to win by 1! This information gain helps you eliminate irrational choices, whereas if you kick first, no choices are eliminated for you.
The actual stats on the two point conversion: "About 47.5 percent of the time since 2015 — almost exactly half that of the extra-point conversion rate." The other point is you have to make the first 2-point conversion to win. You can still tie with the next attempt, but not win. That makes the odds of winning by taking the two point conversion, pre-overtime, is 47.5%. (The odds of losing is higher, because you could also miss the second attempt and not go to overtime.) So, don't go for two.
58 comments
[ 3.5 ms ] story [ 123 ms ] threadAnd when they do, they will be able to capitalize on the information gained from your 2 point attempt.
In terms of expected value, 2 point conversion attempts are roughly par with extra points, though with so few scoring events in the game you would expect large variance.
So outside of scenarios where it is pure upside (1 point doesn't matter without many more scoring events) teams will opt for the lower variance outcome. This matches practice pretty well IME.
If Baltimore special teams doesn't let a punt return TD happen, Houston had one FG to their name... the entire game.
I don't expect a team will always score, absolutely not a team playing against one of the best defenses in football history. So, I respectfully disagree with your thesis. :)
I also suspect that the probability of converting for 2 goes down with each subsequent attempt. You can expect teams to have a few plays installed specifically for critical 2PC attempts, and you may not want to burn those plays for marginal gain early in the game.
these are the assumptions they are attempting to debunk:
- All 2-point conversions are equally likely to succeed.
I have seen no evidence that all 2-point conversions *aren't* equally likely to succeed, and even if they were two bites at that apple gives you better odds than just one
- You will stop the other team from scoring.
regardless of whether you choose to go for it or kick the XP, you still have to stop your opponent on their subsequent drive. if they do so much as kick a FG it's game over already, so it doesn't matter.
- You will get a touchdown on the subsequent drive.
you have to score a TD or the debate doesn't matter so it's irrelevant to this argument
- The clock expires after that.
again we have to assume we are preventing the opponent from scoring any additional points because the debate makes no sense without that caveat
- There are even odds of winning in overtime.
getting to OT in this scenario is the fallback, while it's the best outcome of just kicking the XP twice in a row
(assuming you want to win in regulation which presumably you do because OT is inherent chaotic)
> getting to OT in this scenario is the fallback
I meant to say "likely fallback"
Pass plays are less likely to convert than run plays. The conversion rates aren't even equal between teams.
https://www.teamrankings.com/nfl/stat/two-point-conversion-p...
yes but in this instance we are considering the same team
If you're down 8, with limited time remaining, you should always go for 2 to tie the game, instead of still being down by 1. You'd still need a field goal to take the lead, and that would still be the case if you miss the 2pt conversion.
it is confusing because the conversation is taking place after you just scored
In the linked blog post with the actual statistics in it, she argues that it only increases your chance from 11.2% to 12.9%, so it only really increases your chances by 1.7%. But that just seems to be dishonestly reframing the numbers so people think a 1.7% absolute increase is nothing. But that really is a 15% increase in win chance. It sounds like the data says the 2 points is even better than its proponents claim?
His only other argument is that failing the conversion psychs a team out, which I respect, but that's still just a bare assertion. I don't know why you'd get so mad at statistics for ignoring your feelings. People are complex and absent data otherwise, I would think it just as plausible that the added pressure and imminentness increases performance.
No, that's dishonest. Probabilities are already ratios between magnitudes (desirable vs. undesirable outcomes), so when comparing them you should be subtracting them, not dividing them. A relative increase in probability of 15% doesn't tell you anything if you don't already know the base probability. Also, why go for the 15%? You could also say that the loss probability is reduced by 2%. But I guess you don't get a nice juicy number that way.
I think you're using a restrictive and non-standard interpretation of percentages. They can be used to convey overall probability, but also a change in probability. Only allowing one usage is just a communication deficit.
Not in that particular case, because no one would believe in an increase of 100% (although it would still be preferable to say that it doubles your chances of winning). But if you're at a slot machine and you can put in an extra dollar for a 10% extra chance to win the jackpot, and in fact you chances increase from 0.001% to 0.0011%, would you think that's honest or scummy?
>They can be used to convey overall probability, but also a change in probability.
They can be. They shouldn't be used this way because it lends itself to miscommunication.
Casinos in general are scummy, but I can't fault with the phrasing. What's the threshold at which the world is allowed to use percentages in the way they're meant to be used, without having to be accused of being dishonest?
It's important to distinguish the scummy behavior from the actual statistical practices, as data literacy helps people see through numerical slight of hand.
I think you are backwards from how other people regularly faced with this ambiguity have decided to communicate, given that e.g. finance created a separate term to talk about changes in percentages.
Using that terminology, the win chance has gone up by 160 basis points, which is an increase of 15% from its previous value.
[edit] fixed grammar mistake
> Under completely unrealistic assumptions, G42D8 improves the chances of winning by less 1%. If there is any uncertainty in these calculations at all, then you might even be putting yourself at a disadvantage!
This is implying (without out saying) that a tiny error in e.g. the probabilities of converting a 2 point conversion will make things a disadvantage. After all, what if there's only a 46% (instead of 48% from Ms. Albert's article) of making the 2-point conversion, that must surely lower your win chances by more than 1 percentage point, right?
The answer is "of course not" becuase that just lowers the win probability by a tiny absolute amount as well. In fact this can be seen in the flow-chart that TFA hates so much. The author is either being scummy and disingenuous, or has confused himself by mixing relative and absolute percentage changes.
One number never fully captures the reality. If I have a net worth of 1 penny and I earn a second penny, it is 100% true to say I've doubled my net worth. It also doesn't tell the full story.
If a coach can take an action that would move the win probability from 0.01% to 1%, it would be malpractice to not do so, and correct to say they've increased their win chances by a factor of 100. It's also correct to say they're probably going to lose.
if you fail, you still trail by 8
Or trail by 4, if you really botch it up.
Is this intended to be tongue-in-cheek? Reading such a statement (which logically implies that "if Todd Bowles had used his final timeout, the Bucs would have won") in a post about how other people's reasoning is insufficiently rigorous is off-putting.
Or Bowles mentally marked the game as over and then didn't notice the Lions messed up with the play clock.
> Now, had Bowles used his last timeout, maybe the Lions decide to punt instead of attempting a long field goal. But Bowles surmised a kick through the uprights would have ensued.
Basically, he would have had 36 seconds because of the clock mismanagement on the side of Detroit. Even with 0 timeouts, you can knock out 4-5 solid plays, which is a lot more than 0 solid plays. He goofed up.
https://apnews.com/article/lions-buccaneers-bowles-campbell-...
Claiming that the Bucs lost because of this error is way off, they almost certainly would have lost anyway.
This is misreading the original source [1]. The <1% figure applies when you have not already gotten the first TD.
[1] https://punkrockor.com/2019/11/15/when-should-a-football-tea...
Down by 8 and facing a conversion opportunity means you were down by 14 and just scored a touchdown (6 points). It's more that 1 percentage point change (1.7%) not less (and as others have pointed out, it increases your win probability by about 15%), so there is still a misreading.
- go for 2
- down 8
- convert
Or which sport they come from for that matter.
Some of these I could guess, but 'conversion' is a tricky one.
50% to 62.5% in the simplified calculation is a 25% improvement, not a 12.5% improvement. It's a 25% improvement on already low odds (after all, nobody is actually saying that it's 100% probability that your D will stop the next drive and then you will score another touchdown. It's just assuming that those things have to happen either way so the odds cancel out, and this is looking at win probability assuming that those things happen). Laura's calculations, for a similar scenario (4 possessions remaining), projects the overall probability of winning going from 11.2% -> 12.9%, which is a 15% improvement, roughly the same order of magnitude as the simplified example. I don't have the exact breakdown for a two-possession scenario, but in her slideshow she says this, which seems to be what the author is citing:
> If down by ...16 (before the 6 TD points), the strategy improves the win probability by <1%.
But I am convinced she is talking about absolute percentage points here (e.g. 10% to 10.9%), not a <1% change (e.g. 10% to 10.09%). Why? Because the slide has this before it:
> Rationale: if you are down by 15, you have to score two touchdowns while keeping your opponent from scoring
She is demonstrating that it's not a silver bullet given it's exceedingly unlikely you will win anyway.
I don't understand the point of the article. If G42D8 increases the chance of winning it's the way to go.
Doesn't matter how much the increase is as long it's an increase.
What the probability of losing if you don't G42D8?
Football Clock Management - https://johntreed.com/collections/football-coaching-books/pr...
The Contrarian Edge for Football Offense - https://johntreed.com/collections/football-coaching-books/pr...
Clock management is not just for the last minutes of the game, it is a fundamental part of the strategy for the entirety of play, starting before the game.
I enjoyed them from a not-really-even-a-fan perspective, somewhat like a Moneyball type of read.
The possible choices you can make for 2 PATs are:
Kick, Kick
Kick, Attempt
Attempt, Attempt
Attempt, Kick
If you kick first, let's assume kicking is 100% successful, now your choices are:
Kick made => Attempt
Kick made => Kick again
These are both rational choices; attempting gives you a 50/50 shot to win, and kicking again sends you to OT with a 50/50 chance.
Now let's look at if you attempt first.
Attempt made => Kick
Attempt made => Attempt again (irrational!)
Attempt missed => Kick (irrational!)
Attempt missed => Attempt again
Some of these are irrational! If you made the conversion, just kick it to win by 1! This information gain helps you eliminate irrational choices, whereas if you kick first, no choices are eliminated for you.