I'd like to make a point or raise a question here regarding the magnitude of importance of how bad a trailing team's defense is playing as it relates to the presumed methodologies a coach uses when making a decision to do a, b, or c.
Some sort of combination of broad analytics, instinct, and the application of it all in real time and in real response to what is happening on the field (like using instinct to bump your assessment that Henry is much more likely to convert the 2 pointer in that moment), HAS to include acknowledgement that if a team is down by X number of points with Y amount of time left, 100% of permutations of potential victories involve the trailing team's defense making a stop of some kind at some point. A special teams return could tilt it but that doesn't go into the decision making of when to kick a FG, go for 2, go for the onsides vs kicking away, etc.
If there are 0 winning paths that don't involve your defense making a stop (this would be prior to a game situation where onside kicks are unavoidable), then can it be said that a decision might be tilted one way or another, based on nothing more than faith?
Even if your defense is getting destroyed hard, can you really put on the Dr. Strange Timestone goggles and map out a single path to victory that *doesn't* involve making a stop? If I'm right then I think this kind of goofy faith, where a coach is basically saying "but we *have* to stop them", can explain quite a few decisions that make many of us scratch our heads.
I'm not convinced that analytics are any better than the definitions of the parameters involved. That's not a good or a bad thing it's just a natural limit. I'd be curious to learn how well these analytic models can adjust in real time to game flow context. Like injuries, subs, weather, situations that are anomalous. A coach is *rightfully going* to factor these things in. Hopefully. But the ideal would be to have access to dynamic analytic info in real time and use it, but I'm not convinced it's good enough yet. It's best use for a coach right now is not necessarily those margin calls as much as the ones that are significantly lopsided in favor of decision A vs B.
In poker if you're a 52-48% favorite on a given heads up hand based on starting hands (something like 77 vs AK), you will be infinitely profitable over a long enough timeline to play that everytime. But can you stay solvent long enough to do so? The answer is almost 100% no.
One analytic system says you'll be infinitely rich, and it's true (within the parameter definition). Another system says you'll go broke, and it is also true. What is the problem? The parameters are effed up and there are false presumptions built in. I am not convinced the football analytics folks are doing it right. I think they're really close though and that ultimately they have the right method, which refinement is central to.
Anyway rant over I need to sleep now.