Hi, Thonsew I'd like to correct some factual errors in the message, in respect to Wesnoth AI codebase.
> The cleverness of the AI is limited by time. Right now, in Wesnoth, it is not limited by time. >I have never seen the AI form up on a geographic feature like a river and >force the player to cross. I have never seen the AI use a loose ZOC line of >control to deny some area to the player. True. The reasons are vastly different, however. The AI does not know how to defend well, therefore we force it to attack at uneven odds. Fixing this requires adding new algorithms which will make the AI better at defending. >The AI does not plan multi turn moves. It does, actually, to some extent, for village capturing and strategic movement. >Currently the AI limits the number of candidate actions that is evaluates >based on how long it takes to evaluate a test action. Sorry, but this is wrong. The evaluation of candidate actions is not limited by time. Also, there's no such concept as a 'test action'. >So the t_token change means that the AI will be able to evaluate 25% more >moves in the same amount of time and be more clever. Wrong about the 'more clever' part. -- Cheers, Iurii Chernyi _______________________________________________ Wesnoth-dev mailing list [email protected] https://mail.gna.org/listinfo/wesnoth-dev
