+- Xonotic Forums (https://forums.xonotic.org)
+-- Forum: Creating & Contributing (https://forums.xonotic.org/forumdisplay.php?fid=10)
+--- Forum: Xonotic - Suggestion Box (https://forums.xonotic.org/forumdisplay.php?fid=20)
+--- Thread: [SUGGESTION] Percentiles in statistics (/showthread.php?tid=3053)
Percentiles in statistics - afullo - 05-23-2012
I would suggest to add in player statistics, for the gamemodes in which a Xonotic user has gained sufficient games in order to have a rankable ELO, its percentiles in these modes.
The formula giving percentile of player placed p-th in a list of n players is 100*(1-(p-1)/(n-1)), assigning a percentile of 100 to the first, a percentile of 0 to the last (assuming it exists, i.e. there are no two or more players holding the worst ELO), and decreasing linearly.
For example, a player for which the number of players holding an ELO below its one is three times the number of players holding an ELO above its one, will have a percentile of 75, while a player for which the number of players holding an ELO above its one is three times the number of players holding an ELO below its one, will have a percentile of 25. A player for which there are the same number of players above and below him, will have a percentile of 50.
In this way, it will be more clear how a player is stronger or weaker with respect to other players.
RE: Percentiles in statistics - Mr. Bougo - 05-24-2012
Don't forget to take into account all the inactive players in the system due to unused player keys.
RE: Percentiles in statistics - Antibody - 05-24-2012
I don't see the point in taking a percentile of a number that can already be used to relatively compare players. Rank is based on Elo, and your proposed implementation is based on rank, so what additional value are we getting there?
@MrBougo - I don't think unused keys even come into the picture at this point, given that players need 32 games in order to be considered for ranking and thus this percentile system. Inactive players are a bigger issue, but we can combat that with Elo rotting over time and possibly eventual removal. I think it is safe to say that we don't need to rank players that haven't played in six months, for example
That being said, we don't have enough ranked players at this point for me to put that rot/removal code into place.
RE: Percentiles in statistics - afullo - 05-24-2012
(05-24-2012, 09:02 AM)Antibody Wrote: I don't see the point in taking a percentile of a number that can already be used to relatively compare players. Rank is based on Elo, and your proposed implementation is based on rank, so what additional value are we getting there?
In my opinion, it is more significant to say, for example, "I'm stronger than 70% of players", than "I've got an ELO of 300". So that indicator should clarify the overall strength of a player with respect to all the other players.