This bug has been around for a long time.  If you attack a city that is NOT under siege and there is an army standing on any city tile, the game will treat the attack as if you attacked the army and not the city and the army will be able to retreat. If you win the battle you will not get the city.

Pass the turn then attack the city before the army stading on the city is able to retreat. The army on the city will retreat and die.

Kapaku - Turn 92.sav

I also find it odd that if you don't attack the city the AI will will attack your army, however if you attack the army the AI retreats. If it's just going to retreat from battle, why would the AI attempt an attack?

KnsCtyShufl wrote:

This bug has been around for a long time.  If you attack a city that is NOT under siege and there is an army standing on any city tile, the game will treat the attack as if you attacked the army and not the city and the army will be able to retreat. If you win the battle you will not get the city.

Pass the turn then attack the city before the army stading on the city is able to retreat. The army on the city will retreat and die.

Kapaku - Turn 92.sav

I also find it odd that if you don't attack the city the AI will will attack your army, however if you attack the army the AI retreats. If it's just going to retreat from battle, why would the AI attempt an attack?

We already butted over the first issue, so I'm not gonna say anything ;). 

As for the second, the technical explanation is, that Attack behavior and defensive behavior is governed by two different functions. Which sometimes calculate results that are at odd with each other (even in ELCP to be honest, though it should be rare). In addition to that, the function that evaluates the Military power for the AI is bugged (it thinks it also gains fortification hp if it is attacking, similar to how it works in ELCP). I'll probably write a post about it in my coding bugs thread in the near future.