Originally posted by william78
It's a metric ton - well to do it without a hack in a way that works reliably and meaningfully. Granted Bort's probably one of the most experienced developers on the planet in football games given how long he's been at it so maybe he's got insight to actually solve it - but I used to use this as a question with college graduates after they thought the interview was over (at least those who had a concept of what it is).
The problem isn't the offense, well that's difficult - but it's the defense. The defense can't be perfect it has to be variable but it also has to now recognize (triple option) at least two potential ball recipients who don't yet have the ball and are not currently blocking anyone and recognize that one of those options (the outside) is the most dangerous and that the ball can be pitched/handed off at any time until that person has passed the QB horizontally - and - move with anticipation before finally closing.
That's not even the hard part, the hard part is when you take it another step.
Making AI always react perfectly is relatively easy. The hard part of writing AI is the part where it needs to fail, and do so in a way that's fun and engaging for everyone. Figuring out how they fail, what attributes affect it, and what is a "fair" number in those attributes. What makes it so that it isn't horribly broken at Rookie numbers but also competitive at Veteran level?
As play action has shown, having one AI trick another AI isn't exactly fun. It either works too often to be fun for one team, or doesn't work enough for the other. It's like, yay, we have it...but in practice, what is it? Just a pass play that the defense is designed to react poorly to? Does it feel good to have your defenders react poorly because they were "tricked"?