Does the AI cheat?

[Dave]

I hope to be getting rid of this advantage in my new AI ... at least at a primitive level. A human might still better optimize the movement of leaders and other attack_buffers, but the new AI will at least use it:)

This is a worthwhile goal, and I hope you can achieve it.

The thing that I found really hard when trying to add functionality to the AI leaders was that it's very hard to make them use their leader effectively yet in a way that is sensible enough to avoid the leader being killed, or locked out of its own keep.

The current AI has alot of trouble against 'assasinations' as it is, with its very conservative leader strategy. Using the leader more aggressively would make it even more difficult. Indeed, I think many humans have alot of trouble using their leader yet keeping them safe, the AI would have much more trouble -- and then once a player has worked out how to exploit its flaws, they would be able to do so over and over again.

David

[Darth Fool]

I hope to be getting rid of this advantage in my new AI ... at least at a primitive level. A human might still better optimize the movement of leaders and other attack_buffers, but the new AI will at least use it:)

This is a worthwhile goal, and I hope you can achieve it.

The thing that I found really hard when trying to add functionality to the AI leaders was that it's very hard to make them use their leader effectively yet in a way that is sensible enough to avoid the leader being killed, or locked out of its own keep.

The current AI has alot of trouble against 'assasinations' as it is, with its very conservative leader strategy. Using the leader more aggressively would make it even more difficult. Indeed, I think many humans have alot of trouble using their leader yet keeping them safe, the AI would have much more trouble -- and then once a player has worked out how to exploit its flaws, they would be able to do so over and over again.

David

Well, I was specifically refering to units with leadership, not the "leader" as defined by he 'who you can recruit and you must keep alive.' I agree that the latter problem is a really tough nut to crack. The former problem will be solvable by having the AI able to know that when one of its units is attacking, it should look for units that have been defined as having an attacker_support role (this is not a WML definition) that are within range and move it next to the attacking unit to boost the attack. This will make the AI at least use leadership, even if a human might better choose his attack order to optimize the leadership.