The really low-odds wins have been defensive ones... leaving one damaged unit alone after having taken out a dangerous opposing unit that "absolutely needed to be stopped."
I'm just saying that we as players do a better job of avoiding terrible odds battles. I've won some fights at the minimum odds possible to display though. Saying it "evens" out isn't accurate because a string of far too few outcomes on a statistical basis has a chance of being game-deciding right now.
You yourself has written that RNG-combat resembles deterministic once siege comes into play (if I understand you correctly).
Yes, that is correct. I feel deterministic outcomes would significantly improve the early game while not altering late game strategy much.
Having scouted the surrounding lands well, you would know if you could survive early-game under deterministic combat rules.
You'd need near-perfect espionage on someone to attain that, to the point of seeing what they're producing and what proportion of troops are moving to each location being attacked.
And knowing for sure that you will die if you're boxed in, with Montezuma's huge stack inbound is not anywhere near fun in my opinion. I'll be happy to take my heroic last stand to try to get that 10% win that would secure survival for "one more turn..."
One more turn maybe, but if it's the kind of thing where you have a 10% chance to survive, you aren't going to recover even if you win those battles
. Interestingly, this particular disadvantage is a result of one other random factor (spawns) and deterministic ones otherwise.
Balanced starting-positions? I don't see the fun in that either (except in multiplayer competitive games, that is).
I believe some maps should have it and some not, with some bias towards having them relatively balanced...or at least normalized to the point where you don't auto-lose. There are some starts where against competent players or deity AI you can't do anything at all (3 cities or less, no strategic resources), unless you want to use Inca or Mali 24/7.
On a side note, while I feel these mechanics are too RNG-driven too early, I feel the opposite about the AI; as it stands right now peace treaties, war bribes, and other manipulations make the one thing in the game that absolutely should not be deterministic (other players) far too much so. I would gladly trade more random (though hopefully still viable) choices out of the AI in return for a deterministic combat model. Just as you feel a known combat outcome is dull, I find spamming peace treaties, war bribes, and knowing based on XML that the AI will act as it does is dull. If it were truly a well-written, dynamic AI, I'd be able to look at those XML or even the code and still not be 100% sure at any moment exactly what it would do.
I'll see what I can do about digging up the actual mod; I'll probably need to find it in drop box if it still exists now.
As for the current RNG outcomes, nobody's been able to prove deviation from expected results over large sample sizes except for a rarely-occurring catapult/siege bug that got patched out a while back (basically, the siege would fight an extra round past what it should have under some rare circumstances, but the disiplayed odds didn't factor that extra round).