I'm not sure whether we're seeing anything differently or not. If I read correctly, what you're describing suggests that Evaluate best simulates, well, evaluating: you watch a bit, note the positions/patterns, and then exploit that (such as via Deceptive Attack).

That would be different from suggesting that if you wait a couple seconds, your foe will "drop his guard" a bit and let you exploit that. The latter is how I initially suggested one could use Evaluate + Deceptive Attack as a mechanism for "wait for an opening", as we lack any other mechanism, but I agree that it's an imperfect simulation. Evaluate + Deceptive Attack better simulates what you describe; I think we're of the same mind there.

So how to simulate "wait for and exploit an opening", as a design exercise? I would suggest that the simplest, cleanest method would be to first actually have exploitable openings in the combat simulation. Currently, we can always take a "TH succeeds, AD fails" combination and say, "You found an opening!", but that's not exploitable in the sense that the attacker waits, finds the opening, and then attacks; it's description applied to a generic attack + defense, taken at any time, with waiting playing no role.

If we want fighters to "drop their guard" involuntarily (who would do it on purpose? : ) at specific moments, I'd think we'd need to randomize defenses a bit before any TH or AD rolls, in a manner that attackers can attempt to observe and then act upon. Or let Evaluation allow some roll to see (roughly) what the AD roll will be before the attacker decides whether to go for it. Something like that. (But whatever a solid simulation may turn out to be, any extra mods or rolls will likely limit its use to one-on-one combat.)

Incidentally, I love open-ended scaling too, applied appropriately. I prefer your version of Evaluate, and have considered the same – though we have to consider how far to take it. Evaluate indefinitely, and the bonus will go up forever... Should it be capped somehow? Or can the opposing Evaluations of two circling guys do something to "break" the other guy's accumulated bonus? Interesting potential there...

But while an open-ended bonus on something like TH or Feint is potentially problematic, an open-ended bonus on noticing something is much less so, IMO. So if there were, for example, some roll to merely notice when a foe's guard is weak (whether that's simulated through random mods on AD, or "pre-knowning" his AD roll, or some other mechanism), with an attack based on that successful observation gaining some nice (but not open-endedly infinite) bonus, then an open-ended bonus applied to the observation roll may work nicely.

Again, tossing it out to see what sticks –