| |
BGonline.org Forums
VRR - Challenge
Posted By: Bob Koca In Response To: VRR - Challenge (Rick Janowski)
Date: Friday, 30 January 2015, at 9:12 a.m.
1. XG gives VRR automatically. There is a problem in my mind though in that there is a contraction away from extremes of 0% and 100% performed. This can be undone though. I disagree that the contraction should be done.
2. I think an Elo system using VRR results would be ideal. For example if Elo difference predicts a 60% win chance then one would need to be have a VRR at that level to have a zero change in rating. The greater the difference from that level the greater the rating change.
3. To get a rough idea of the effectiveness consider the difference of an opening position RO with and without VR. Roughly each game with VR is worth about 70 without.
Regarding reliability suppose A is a better player than B and they play 50 seven point matches. Which is more likely, A having a lower PR or A having a higher total of VRR? I am not sure. My guess is that they are close but many seem to think PR does better. I don't have a good argument why and I don't think anyone has given a good explanation one way or the other.
The VRR crowd wants everyone to realize that the PR approach measures who can play with a better PR and VRR approach measures who is better at winning backgammon matches and there is not a perfect match in those goals.
For purposes of your BMAB title giving, the use of VR has a big problem of measuring difference against a certain population of players but not giving a number like PR does that can be compared fairly. A strong set of Variance reduced results may just be due to a weak set of competitors. I see it as preferable in a dual duel type system. Instead of a PR winner look at VRR winner. Or even better, have a Round Robin and add the VRR which does not throw away information about size of a victory.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.