Suicide attacks

[Ninja2]

What the purpose is of this sort of behaviour, I'll never understand. Typically, I'm in a limited type of war with the AI. The AI will drive a stack up to one of my cities. When the AI originally sent the stack, it was strong enough to capture my city, which might have had just one or two defenders, no walls and a weak cultural defense. Your standard border city. At this point, the AI will use it's one or two siege units to reduce my cultural defenses over the course of several turns, which gives me more than enough time to reinforce and build (whip) walls, and otherwise start mobilizing a counterattacking force. The AI is now severely outnumbered (but might not be directly threatened by my city defenders), but in some cases it will burn its stack in suicide attacks. Of course, a much better approach would be to either 1) reinforce the attack stack with more attackers and collateral damage units, 2) withdraw the units if I bring in more offensive units and prepare a counterattack or 3) bring in counters to my offensive units and simply maintain the threat to my city, while possibly attacking in another location.

Is there a way to make the AI stop the suicide attacks?

 

[Yakk]

Yes. It needs a better idea of if it's attack will work or not.

Currently, the AI is really bad at knowing if it's attack is pointless or not. And has little to no knowledge of "there is another stack also next to the city who will join in the attack", and the like.

The combat simulation thread is looking into a way to give the AI stack-on-stack prediction abilities (which can then be extended to local-forces on target-stack prediction abilities). Then the AI can say "if I attack, I'm screwed -- so withdraw, or bring up more forces".

 

[jdog5000]

Or it could continue bombarding ... I hope the stack assessment will help the AI make lightening strikes, and not stall bombarding (and letting more defenders in) if it would likely win without an extra turn of bombards.

 

[Elenhil]

With vanilla AI I had a wonderful time circumventing a siedge of my city. The enemy was across the river bombarding my 120% defence city, but there were roads (and bridges) across the river to the squares neighbouring his stack. So I could move my attack units across the river, strike his stack without the river penalty and retreat to the city to heal up even with the slowest of my units thanks to those roads. If only the AI could see the necessity of destroying these two roads, I would have had no opportunity for such strike-and-retreat moves, quite possibly loosing the siege.

Can BetterAI handle such logistical planning as to destroy routes of possible counterstrikes?

Also, the city was more than one move away even for my knight units. Which meant its garrison could not be resupplied immediately. Which meant just one more strong stack on my side of the river could destroy all the reinforcements one by one. So a counter-reinforcements planning (if possible) would be a marvellous improvement.

 

[Yakk]

No, the AI is far away from that level of stategic understanding of the map.