Upgrade your account to a Community Supporter account and remove most of the site ads.

Reply to thread

Message: <blockquote data-quote="FormerlyHemlock" data-source="post: 6939034" data-attributes="member: 6787650">Has anyone here ever read Pieter Spronck's papers on adaptive AI difficulty in games? You can set up a script of rules for an AI to follow, where some rules are more effective than others; and then as players get better at beating the game, instead of increasing the AI's power (extra resources, cheating, etc.), you let it unlock extra rules and get smarter.The D&D equivalent of this would be things like: against new players, Minotaurs always Reckless Attack. Against skilled players, Black Minotaurs (same game stats but more cunning) cunningly calculate (or at least approximate) the optimum time to Reckless Attack (i.e. only against heavily-armored foes). Against new players, bandits just shoot whoever is closest. Against skilled players, Elite Stormtroopers (same stats as bandits but better-trained) seek cover, drop prone against missile fire on their turns, and engage spellcasters at long range (250'+) while spreading fire out enough to force inefficient expenditure of Shield spells. Against new players, a T-Rex just attacks and starts munching a PC. Against skilled players, an Awakened T-Rex with Int 10 will use its superior speed (50') and reach to grab one PC at a time and munch him to death in its jaws while avoiding the other PCs. It will take opportunity attacks if necessary in order to avoid a full Extra Attack sequence. It may target backline squishies preferentially if it can determine their threat level based on their behavior.</blockquote>

Verification