One rest, after 5 combats, allowing fighter to dump full superiority dice on the last combat. It's orthogonal to my goals to work out the optimum.
Exactly. Our wonderful numbers on paper fail the rule that For damage to be effective, it must be applied. (Consider the contrary, for damage to be effective, it must not be applied!?) As you point out, GFB and Sentinel require a party, and require foes to be arranged in just the right way. [MENTION=6785802]guachi[/MENTION]'s GFB builds has the same problem. My belief is that all our estimation is empty until we come up with a construct that factors in applicability.
accuracy = lesser of 0.95 and (20−(AC−attack))*0.05
average = ((dice minimum)+(dice maximum)/2)+(flat modifiers)
damage = average*accuracy
cleave = 1−((1−(damage/HP))^(number of attacks))
critical = 1−((1−critical chance)^(number of attacks))
depletion = expenditure/turn*5
short resource fraction = charges/2
long resource fraction = charges/6
We're familiar with most of those and apply them with varying consistency, but what about
space = distance−range ?
lag = space/speed ?
awareness = binary assumption? ratio?
Without those, along with foe count and positioning assumptions, and ally count and positioning assumptions, our estimates frequently yield intuitions that play proves false.