| 1 |
[quote]But the random swapping might increase elo variance again, potentially to the point where it's back to the old problematic cases.[/quote]
|
1 |
[quote]But the random swapping might increase elo variance again, potentially to the point where it's back to the old problematic cases.[/quote]
|
| 2 |
I tried a very simple descent algorithm (take the smallest step in the right direction possible, stop if there is no such step) on random data (continuous distribution, might not represent real elo data I suppose) and found that for medium-to-large problems (where the original solution isn't good enough, which is itself rare for large problems) it converges to a good-enough solution quite rapidly, without disturbing the elo variance much.
|
2 |
I tried a very simple descent algorithm (take the smallest step in the right direction possible, stop if there is no such step) on random data (continuous distribution, might not represent real elo data I suppose) and found that for medium-to-large problems (where the original solution isn't good enough, which is itself rare for large problems) it converges to a good-enough solution quite rapidly, without disturbing the elo variance much.
|
| 3 |
\n
|
3 |
\n
|
| 4 |
In general this depends on how much random swapping one permits oneself.
|
4 |
In general this depends on how much random swapping one permits oneself.
|
| 5 |
\n
|
5 |
\n
|
| 6 |
[quote]That issue is this: even assuming that !predict is correct and that the teams as matched by the balancer both have close-to-equal chances to win, the balancer sometimes produces matches which seem unfair, where "unfair" here means not merely a lopsided chance to win, but perhaps more importantly a lopsided chance to enjoy the game due to one side having a much higher number of much lower-skilled players.
|
6 |
[quote]That issue is this: even assuming that !predict is correct and that the teams as matched by the balancer both have close-to-equal chances to win, the balancer sometimes produces matches which seem unfair, where "unfair" here means not merely a lopsided chance to win, but perhaps more importantly a lopsided chance to enjoy the game due to one side having a much higher number of much lower-skilled players.
|
| 7 |
[/quote]
|
7 |
[/quote]
|
| 8 |
I don't know that there's even a discussion to be had here. I am not sure that one-team-has-high-elo-variance games are *imbalanced* (and if they were I don't know which way) but I am sure that they are *unfun* and to be avoided where possible. What else is there to say?
|
8 |
I don't know that there's even a discussion to be had here. I am not sure that one-team-has-high-elo-variance games are *imbalanced* (and if they were I don't know which way) but I am sure that they are *unfun* and to be avoided where possible. What else is there to say?
|
|
|
9 |
\n
|
|
|
10 |
The difficult part is how to accomplish that without compromising anything else.
|
Balancer broken?