Loading...
  OR  Zero-K Name:    Password:   

Post edit history

ELOs again, how about some ultracompetition?

To display differences between versions, select one or more edits in the list using checkboxes and click "diff selected"
Post edit history
Date Editor Before After
2/28/2017 12:08:04 PMDErankBrackman before revert after revert
2/28/2017 11:56:39 AMDErankBrackman before revert after revert
Before After
1 Indeed elo is not completely correct. I just mean, the less it is correct, the worse is picking. If the actual elo is also used in the probability prediction part of elo change, then this part is ok. 1 Indeed elo is not completely correct. I just mean, the less it is correct, the worse is picking. If the actual elo is also used in the probability prediction part of elo change, then this part is ok.
2 \n 2 \n
3 In principle, percentiles are good, especially if they are equally large. Then the belonging to a certain percentile provides most information. The bad thing is that you don't know for which numbers they stand. Maybe it should be made so that the elo number limits of percentiles are only reset every month and made public. 3 In principle, percentiles are good, especially if they are equally large. Then the belonging to a certain percentile provides most information. The bad thing is that you don't know for which numbers they stand. Maybe it should be made so that the elo number limits of percentiles are only reset every month and made public. Or they could be reset once per year or never, and rounded to 50s. They won't change much anyway.
4 \n 4 \n
5 I think we all agree that decayed elo is not a good measurement of skill, its only purpose is to promote activity. Why is decayed elo used for balancing and matchmaking then? 5 I think we all agree that decayed elo is not a good measurement of skill, its only purpose is to promote activity. Why is decayed elo used for balancing and matchmaking then?
6 If we want to reward activity, I have a better idea: Atm, the rank icon is influenced by how much you have played in the past and how good you are currently (+how much you play currently). Instead, we could make it depend on how much you play currently and how good you are currently. So the color would be determined by elo without any decay and the form by how much you played in the last weeks or months (or the other way round). 6 If we want to reward activity, I have a better idea: Atm, the rank icon is influenced by how much you have played in the past and how good you are currently (+how much you play currently). Instead, we could make it depend on how much you play currently and how good you are currently. So the color would be determined by elo without any decay and the form by how much you played in the last weeks or months (or the other way round).
7 \n 7 \n
8 The kind of decay in @DeinFreund's WHR is much better. Here, deviation increases, not malus. Even in ZK's current rating system, weighting could be decreased instead of malus increased. 8 The kind of decay in @DeinFreund's WHR is much better. Here, deviation increases, not malus. Even in ZK's current rating system, weighting could be decreased instead of malus increased.