Next gen AI ?s

Chalnoth, I remember adjusting an actor for ctf levels in the original ut, which was "alternate paths". So the bots could "sometimes" pick these alternate routes to go back and forth to enemy base.

Most maps, including the retails ones weren't bothered to include this feature, but in the end I came up with interesting results. Although this actor was something static. It was just a probability measurement tool for the bot and it didn't change during the gameplay. I don't know it would be possible to alter such a dynamic version of it to tell the bot if such an alternative path would be useless/full for a given particular time during the game play. So the bot himself could surpass the rigidness of one probability. I don't also know if this relates to "true learning" factor in ai.

Anyway, the bots in ctf in original ut still amazes me. Afterall, it was just a matter of how they perceived the bsp for its "easy-to-navigate" property and always pick the wider road rather than narrow one. Strangely enough, while I spectated the bots, they were always prefering spacious places rather than cramped ones even the goal at the end was identically the same. If there were two corridors leading to one big room and one corridor was 256 units wide and the other one was 512, the bots always prefered the 512 unit corridor to lead into the big room.
---

Thanks for the info 3dilettante, it's always pleasing to get some more info about game ai

 

do you think they'll be able to make some bots dumb and some smart? cause if all bots 'learn' too well, no one will ever beat another game again! i hope they put a 'dial' on it. cause us non 'killer' types won't be feeling too 'uber' (like we like occasionally) playing against a bunch of predatory genius bots.

 

Well, you can have bots that both have good strategy and are easy to kill. The two aren't really connected that directly, and bots that have good strategy will feel more organic.