I have an issue with the probabilities that are being used. This isn't a strict mathematical situation that has no outside influences. It's not like a roll of a dice or a flip of coin that all participants are equal and match the probabilities. Every team is different. Offenses are better or worse as are the defenses. That particular game may have a weakness found because of an in game injury making it harder or easier to succeed in a two point conversion. There are way too many variables to take strict historical data of every NFL team going for two to calculate a win probability if you go for two or if you don't go for to. All things aren't equal which is a huge assumption when trying to argue for or against this situation. There is no one size fits all.
I'm not buying this at all. Just about any situation outside of a lab has outside influences. Yes, every case is different. That's the whole idea behind using
averages. By your logic, teams should never bother using any sort of quantitative analysis at all, because every situation is unique. "Should we go for it on 4th and inches? Well, there's no way to tell, because the current wind speed is 2MPH faster than average, so we have to throw out the entire data set."
The fact is, every day across different industries, people make quantitative decisions based on incomplete data. Financial analysts try to predict stock price movements. Fivethirtyeight.com is attempting to predict every
House,
Senate and
governor's race, which is an act of pure hubris. Will they get plenty of things wrong? Of course (assuming, that is, that you believe the occurrence of a result to which a model assigned a 30% probability counts as being "wrong"). But do those models tell us more than we might know from pure guessing? Generally speaking, yes.
Going for two in this situation won't always turn out well. And sure, there may be scenarios where it makes less sense. But
on average, teams that do it will fare better than teams that kick the XP. And when you consider that, prior to two weeks ago, there had only been one instance
in this century of teams going for two there, that suggests that the persistence of XP attempts in that scenario is not because coaches were carefully considering variables unique to each situation, but rather because certain biases were causing them to make sub-optimal decisions.
Let's turn this around: Can anyone make the positive case for kicking the XP? Because throughout this entire thread, all I've heard is "something-something-take-the-points-blah-blah-can't-trust-math-durrrrr-momentum." None of those arguments offer any evidence, quantitatively or qualitatively, that kicking the XP makes you more likely to win the game. Absent that, and given the presence of an estimate suggesting you should go for two, what rationale is there for kicking?
If this was about playing craps and rolling dice then it would make sense to say always do X because it gives you a better chance at Y.
If you're applying quantitative analysis to craps and haven't figured out that the house always wins, you should get out of the casino before you lose your kids' college funds.