I should probably stay out of this, but fools rush in where
What I think I’m hearing from you and @HappyDays seems to me like the somewhat different statistical conclusions reached by Bayesian and non-Bayesian statistics. It’s analogous to the “Monty Hall Problem”. The classical statistics approach is to look at the set conditions and calculate a generally applicable probability. The Bayesian statistical approach is to utilize additional information available to modify analysis. In the case of the “4th Down Bot” the information would include how successful the specific IOL is at the play, how successful the DL is at stuffing it, any other available details applicable to that given situation with those specific teams.
JMO
But if you look at the link above, you’ll see that the Monty Hall Problem was hotly contested by very knowledgeable smart people, both ways