Vengeance
I am not clear in how does anything that you stated contradict what I have stated in this?
I think you are assuming vast variations in how the monster acts each time, I did not mean it in that way. I meant that you don?t want it so that (eq example) you pull the orc from tent 2 you know that the orcs in tent 1 and 3 won?t come over to beat on you. The AI should leave a bit more chance then that.
Creatures have character traits, they do X a lot more often then Y, and they never do Z.
They will also have weaknesses and foolish things that they almost always do, like their main caster may be really good at heals, but he will only heal at the end of combat, or when noone is attacking him, and his friends won?t defend him.
In this case a well rounded group of monsters or MOBs are actually less then the sum of their parts, players learn to have one player attack the healer and then they no longer must worry about him healing.
I never stated anything about making the ultimate challenge to the players.
I am almost positive that we agree on this, but I could be wrong.
In how the programmer intends the AI to be is where we seem to have a difference, I am assuming he wants the creatures to be beatable, but if that is not the intent, well that is his prerogative. In many single player games, part of the puzzle of a certain part is that you can not defeat something in combat, so you must do something else. In persistent worlds this is more complex and not as desirable, as it is more annoying to die 20 times then it is to reload a saved game 20 times, until you learn your lesson.
But tell me if we still have a difference here :-)
(this is a great example of the dev team in discussion mode, arguing over something we are not sure we agree on and having to explain the point to verify if we actually disagree or not)