Quote Originally Posted by Astarica View Post
People seem to have this mistaken notion that AI is about the computer trying to win. It's never the case because the computer can pretty much win every single time in almost any game if it actually tried. It doesn't even have to cheat most of the time, but of course it can do that too. Randomness is not smart. If you're controlling Titan, and suppose you have to obey all the aggro/CD/targetting rules and you're playing against 8 really good players. After a while you'd notice Landslide isn't actually hitting any of those guys and you'd stop wasting your time on that. You'd likely settle on a rotation like Bomb -> Tumult -> WotL -> MB (to maximize the amount the healer have to move before you try to finish tank off with MB), and if this rotation works you'd just stick to it 100% of the time. There would be nothing random about it, and there's no reason for you to do anything different once you found a rotation that can defeat the players.
That's not actually entirely true, though it does get at the heart of what I'm trying to say. video game (Chess aside) AI in general does not follow true AI principles, because if it did, it would destroy all challengers. That being said, if you used a min/max tree as you basis, chances are pretty good you actually wouldn't end up with an AI repeating the same actions again and again. The reason for this is because the AI is always reassessing the validity of a given move based on the current environment. If the players try something new to counteract the rotation you mentioned, the min/max tree would fold that into consideration and change it's tactics accordingly.

I suppose my point, since it seems to have a lot of support, is that you can't just make something more challenging, be it by randomizing it, or using more and more complex patterns, because there comes a point where you have gone too far and it is no longer the experience the players came to play. To make it more complex, each player has a different tolerance point.