No replies to this topic

[LordNothing]

the recent event has hilighted what has been an issue for a very long time. random map selection is not good enough. despite on paper randomness people are still drawing map streaks or not seeing maps they havent played in awhile. i remember when forge came out i had to play cw continuously for a week to see the new map. i think you can do better.

instead use heuristic analysis to identify what maps that a group of 24 players would like to see based on their recent map history. give each map a rating and use that rating to weight maps based on how much or how little playtime these 24 people have put into them. then use those weights to make random selection favor maps that those players have not seen in awhile.

then it comes down to deciding how to weight maps. the most basic idea is to list all players last 10 maps and simply count them. maps with lower numbers would get the higher selection probability. though i have a feeling that would result in map selection not much better than we have now.

if we wanted to break map streaks, we could also take each players list and count up exponentially, so that if a map has appeared a number of times in the last game, it would count up exponentially. then the totals for each map would be summed up across all players' histories.

if you wanted to account for time since the map was last seen you can divide each map by its place in the history so that maps that have been played most recently have higher values than maps that were played 5 games ago.

this should make it seem as if the maps are rotating from each players perspective when in fact they are still being selected randomly. but its also better than an actual rotation because you avoid the issue of getting synced up to the cycle.