SO something happened with Patch Z and AIs....

[Emptiness]

Yes, the AI as a paradigm of efficiency never does anything that harms itself. What could my words be if not paranoia? If only there was a way for people to observe the behavior I described for themselves...

 

[Humakty]

@ Emptiness : I've spotted too this behavior, it is obvious to me the AIs know were to strike and when. They most of the time go for my less defended city, even if it is hidden from them all game long (as in : they can't see it, nor what units defend it)

This is why I always accept open borders with AIs, they won't learn anything they don't already know

 

[Lone Wolf]

I've seen the AI's attack my heavily fortified border cities instead of lightly fortified border cities dozens of times.

 

[Humakty]

I wrote 'most of the time' because often they go for the city which is the nearest to their borders. I think the AIs tend to calculate the odds before taking a decision. If the border city is too well defended, they'll ignore it.

My personnal example (the one that springs to my mind) is a border city with ten well promoted units in it, and another city 4 tiles southward with one unit defending it. They aimed right to this city, ignoring the well defended one. I must say the other city was surrounded by mountains, and they couldn't have scouted it.

You must not think I hate this game because the AI knows stuff she should not, I think it is a part of it, and a necessary evil to render the AI at least equivalent to, err, a parrot.

So a basic trick is to leave equivalent but very mobile garrisons (which can regroup in the attacked city in one turn) in your cities, so that the AIs will aim for the closest one. Et hop ! You know were their attacks will land. It's magic ! (you can also leave big stacks wandering in the country, as they don't seem to be taken in account)

 

[Akasha Valas]

I wrote 'most of the time' because often they go for the city which is the nearest to their borders. I think the AIs tend to calculate the odds before taking a decision. If the border city is too well defended, they'll ignore it.

I'm not 100% sure on this, but I've scanned the code where the AI picks its targets before and it doesn't seem to actually do this, which would explain why it likes throwing units at Acheron when it has no chance to win (even after I dramatically decreased the AI's internal odds calculation for any unit without courage vs. any unit with fear, it still did it).

 

[Lone Wolf]

My personnal example (the one that springs to my mind) is a border city with ten well promoted units in it, and another city 4 tiles southward with one unit defending it. They aimed right to this city, ignoring the well defended one. I must say the other city was surrounded by mountains, and they couldn't have scouted it.

Never seen such a thing. They just suicide stuff on cities.

 

[avalonnn]

Maybe some of the 'cheating' happens at higher difficulty levels?

 

[kenken244]

No, it happens in every difficulty level. The problem with not allowing it to see things is that the AI can't look for things either. for example, if it wiped out all cities of a civ that it could see, it would not be able to attack the city just outside of it's mapped area. Humans are smart enough to look for the other cities, but to the AI they just don't exist.

 

[deadliver]

Most of that is paranoia.

Actually, if the AI behaved that way, it would harm them more then it would benefit them.

Also that is easily countered by doing that same to the AI, which I do. If you can cut them off early, they will be weaker. Even if they DoW you probably just have to fend off their all or nothing attack stack and mop up the stragglers.

 

[Monkeyfinger]

There are different preset AI scripts for individual units, and what script a unit has determines whether or not it'll attack a city that's too well defended to be cracked by the stack it's in.