A new League of Legends devblog offers an interesting read for any MOBA players wondering about why bots act like they do.
According to the devblog, Riot surveyed a bunch of players about bots (as botmatches make up a large part of the total League of Legends playtime, even at high level) and learned that one of the things most requested is that bots act more “human.” To that end, the devblog discusses improvements made to the bot AI, how they were made, and why some improvements are really, really difficult to implement.
It’s a genuinely interesting look at bot AI, particularly in a game like this. For instance: there’s still no jungling or wall-jumping because of the way pathing works – bots don’t really “see” walls, per se, instead following paths that they’ve been given. Jungling relies a lot on picking safe routes rather than just getting from A to B as fast as possible, so it relies on entirely different thought processes than just moving around.
However, they should now attempt to dodge skillshots (which was a problem before because, again, they didn’t actually see skillshots), they should be better at judging when to go into a fight or not (as before they mostly just looked to see how much damage they’d sustained and used that to determine whether to stay or run… which doesn’t make much sense in a game where lots of high-damage abilities also have high cooldowns), they should use combos and builds that are currently part of the meta, and a whole lot of other stuff.
Regardless of whether or not you play League of Legends, it’s a fascinating read into how bot AI works. Give it a look over here.