The issue with increasing 'ATTACK_CITY_STACK_RATIO' and things like that is that although it may solve your current issue, it will make the AI more likely to sit around doing nothing when their stack is already big enough to take a city.
A ratio of "100" in this context means that the AI thinks its stack is equally strong at attack as the enemy is at defending. A ratio of anything greater than 100 means the AI thinks its stack should win.
The calculation of the ratio isn't exact and so it isn't always correct. It actually doesn't properly account for promotions which give bonuses against particular units (such as 'Charge'), but it does take most promotions into account. It's difficult to calculate the effect of things like Charge, because it depends on which particular unit faces up against which other unit - which depends on the order of attacks, and on the outcome of individual battles, and stuff like that.
Players generally don't want to "barely" win a stack attack, they want to win comfortably so that they minimise their loses. That's the main reason for wanting a higher threshold for the attack ratio. But really, if the AI attacks with a ratio of 120 and loses, then that's either really bad luck, or a flaw in the calculation. I don't think it should be corrected by increasing the attack threshold. If it really is a problem, it should be somehow addressed by making the calculation more accurate; otherwise we'll just end up messing it up for all the other cases where the calculation is already accurate.
Horatius is right that there are some personality modifications involved. And it isn't only what's in the xml files. I've personally written some stuff which adjusts the attack threshold based on AI personality and on the particular situation and strategy. So it can vary quite a bit, and it isn't surprising to me that sometimes a stack attack fails. I think it's important to remember that sometimes human players make that same mistake!
--
Incidentally, they'll only continue preparing for 20 turns (on 'normal' game speed). If you still want them to keep preparing after that, you'll have to remind them. (Or just start the war. They don't need to be told when they are actually at war!) 20 turns isn't really a huge amount of time, but the AI doesn't trust that you'll remember to tell them to cancel the war preparations if you change your mind... Maybe the AI should contact the human to ask if the war is still on rather than just silently cancelling the plans; but I don't think it matters much.
--
By the way, the main reason the new patch doesn't have more balance changes is that I haven't been in a position to play-test properly, and I don't feel strongly enough to just push through any particular changes right now. In any case though, xml changes are something that individual players can pretty easily make themselves to try out their own ideas. So I figure it doesn't matter much if I just leave that alone for the time being.
A ratio of "100" in this context means that the AI thinks its stack is equally strong at attack as the enemy is at defending. A ratio of anything greater than 100 means the AI thinks its stack should win.
The calculation of the ratio isn't exact and so it isn't always correct. It actually doesn't properly account for promotions which give bonuses against particular units (such as 'Charge'), but it does take most promotions into account. It's difficult to calculate the effect of things like Charge, because it depends on which particular unit faces up against which other unit - which depends on the order of attacks, and on the outcome of individual battles, and stuff like that.
Players generally don't want to "barely" win a stack attack, they want to win comfortably so that they minimise their loses. That's the main reason for wanting a higher threshold for the attack ratio. But really, if the AI attacks with a ratio of 120 and loses, then that's either really bad luck, or a flaw in the calculation. I don't think it should be corrected by increasing the attack threshold. If it really is a problem, it should be somehow addressed by making the calculation more accurate; otherwise we'll just end up messing it up for all the other cases where the calculation is already accurate.
Horatius is right that there are some personality modifications involved. And it isn't only what's in the xml files. I've personally written some stuff which adjusts the attack threshold based on AI personality and on the particular situation and strategy. So it can vary quite a bit, and it isn't surprising to me that sometimes a stack attack fails. I think it's important to remember that sometimes human players make that same mistake!
--
Telling your ally to prepare for war should make them start building whatever they would have normally built if they were preparing for war themselves - including ships and transports.Does anyone know if the "please prepare for war with civ X" diplomacy command makes the AI make more ships and transports to be able to attack island civs or does the Ai only make more land units?
Incidentally, they'll only continue preparing for 20 turns (on 'normal' game speed). If you still want them to keep preparing after that, you'll have to remind them. (Or just start the war. They don't need to be told when they are actually at war!) 20 turns isn't really a huge amount of time, but the AI doesn't trust that you'll remember to tell them to cancel the war preparations if you change your mind... Maybe the AI should contact the human to ask if the war is still on rather than just silently cancelling the plans; but I don't think it matters much.
--
By the way, the main reason the new patch doesn't have more balance changes is that I haven't been in a position to play-test properly, and I don't feel strongly enough to just push through any particular changes right now. In any case though, xml changes are something that individual players can pretty easily make themselves to try out their own ideas. So I figure it doesn't matter much if I just leave that alone for the time being.