Here is a great example of a flawed 4th down analysis. It gets too in the weeds by thinking too hard about the EP of your opponents starting field position.
Their argument if I can summarize is:
EP(going) = EP(convert) - EP(fail)
EP(fieldgoal) = EP(fg_good) - EP(ensuing_kickoff) - EP(fg_missed)
And the rule is if EP(going) > EP(fieldgoal) then go for it.
BUT what they miss is that EP(going) ≠ EP(convert) - EP(fail) because all of those EP(convert) situations are going to result in situations where your opponent will subsequently receive the ball. Because it's a terminal play for the drive they account for this in the EP(fieldgoal) = EP(fg_good) -
EP(ensuing_kickoff) - EP(fg_missed). To make the analysis comparable they need something like:
EP(going) = EP(convert) - EP(ensuing_possession_successful_conversion) - EP(fail)
EP(fieldgoal) = EP(fg_good) - EP(ensuing_kickoff) - EP(fg_missed)
where EP(ensuing_possession_successful_conversion) = βEP(ensuing_kickoff) + (1-β)EP(expected_fieldposition), β, odds of scoring drive ---- or some garbage like that.
imho unless you're getting paid for the analysis just cancel all that subsequent possession **** out of each side of the equation to come up with a rough equilibrium like I did. It's at worst a marginally bad decision to go for the FG there and probably correct, imho.