Anyways it seems my ai does terrible against yours.
Probably. I don't know what's going on, the difference in power is so obvious, that one would expect > 90% victories for you. Instead, it's only 60%. Of course, I don't test with Aztecs, but neither with Koreans. I prefer civs with some infantry, because cavalry/archers are countered rather easily by Byz.
Actually they should change it knights to militia if they either have more food than gold in late castle age or when they're advancing to imperial. Probably the vikings were already dead by then so you haven't noticed that.
They changed the army, but a little too late imo. They only started training something else but knights by 30'. Instead of 35 knights, they could have 15 knights, 20 militia and 20 archers with the same money.
Rams are correctly trained late, but I believe that you should block them (maybe have a couple as arrow magnets) when you have lost your advandage.
As for the bad dark age in that game, probably the reason was berries. It didn't lose villagers to wolves, maybe only one (by viewing the chart in the end). I think that this 1 villager difference was due to loom (later it got lost when Lade loomed too).
I only highlighted the dark age because I believe that it was some bad luck the reason that those 2 games were so much different. So impressive in the first, rather pale in the 2nd, same civ/map/opponent. Simply a matter of luck, but maybe the ability to adopt played some role. It adopts well in strange situations, but maybe you could boost it a little more.
IMO, cheap units help adopting. ATE, VNS_Halen and No Limits use them pretty well, imo.