What about collecting suggestions for AI improvements in one thread? At this stage I suppose it might be better to stick to simple rules rather than complicated strategies, but the devs might step in on that…
Here’s one situation: AI ship 2/4/0 attacking a village 5/6/1. Neither upgrades. Result: the village loses 1 health, the AI ship loses 4 and is sunk.
Proposed rule: the AI should not decide to attack if that will surely result in sinking its own ship.
That’s right. I add that the AI ships never attack one or more ships next to one settlement, even if they’re weaker and they can be sunk in one turn.
The AI shouldn’t attack a target that they won’t cause any damage to, or that will result in their own death. But if they can’t find anything better to do, this still might happen occasionally.
Doing nothing is better than causing your own death
I have to say that AI taking over after one missed turn (the way AI) currently operates is not really working.
The AI will settle anywhere, next to my own settlement where it collects a food tile and a sea food tile… That’s it…
It also attacks neighbours, a fragile truce ruined when the AI attacked a trusting fellow pirate
Most problematic is the sending of exotic resources to other pirates! It’s not to all pirates so some get a significant bonus whilst others get nothing. Not mentioning the loss of the resources when you rejoin.
In the next update, the AI will only start making decisions after you’ve missed more than 1 turn @thyrus.
The AI assigns a hefty penalty for attacking if it will die or can’t cause any damage, so it is unlikely, but still possibly in edge cases, @ruisselet.
I think clear trading rules would be useful. What currently happens if you send AI a resource? Also, are there steps you can take to stay friends with AI?
Can I also suggest that if the AI is friends with all it’s immediate neighbours for a certain number of turns (say 3 consecutive) then it can double-cross someone and turn on them (maybe the strongest?)
Thanks for the answer. It still happens though. Why not filter out such bad actions entirely?