Currently
- a beginner map only played by beginners and beat by 50% of them
- an expert map only played by experts and beat by 50% of them
will get the same map rating. I think this is not desired behavior because it causes people to find maps that are not the right skill level for them.
I suggest a hidden elo system, where both the map and the player start at 1000. The map's elo is essentially a rating of a player who has a 1/2 chance of beating the map.
If a weak player beats a hard map, then the map rating is decreased and the player's rating is increased.
If a strong player beats an easy map, there will be barely any change because they will have been predicted to beat it.
This doesn't need any new data, and it's way more accurate than raw completion stats. the only thing I am still unsure about is 1. what should count as a play, because I got this comment on Discord:
that map is obstaslop, maybe a lot of people joined and left because they don't wanna play that, even for 40s
And 2. if completion rate should be the only metric, maybe length/playtime can also be taken into account, but it could discriminate against one jumps or other short-but-hard maps (I joined the server and couldnt find any one jump maps so not sure if this is even a thing anymore)
Currently
will get the same map rating. I think this is not desired behavior because it causes people to find maps that are not the right skill level for them.
I suggest a hidden elo system, where both the map and the player start at 1000. The map's elo is essentially a rating of a player who has a 1/2 chance of beating the map.
If a weak player beats a hard map, then the map rating is decreased and the player's rating is increased.
If a strong player beats an easy map, there will be barely any change because they will have been predicted to beat it.
This doesn't need any new data, and it's way more accurate than raw completion stats. the only thing I am still unsure about is 1. what should count as a play, because I got this comment on Discord:
And 2. if completion rate should be the only metric, maybe length/playtime can also be taken into account, but it could discriminate against one jumps or other short-but-hard maps (I joined the server and couldnt find any one jump maps so not sure if this is even a thing anymore)