
BGonline.org Forums
VRR  Challenge
Posted By: Rick Janowski
Date: Friday, 30 January 2015, at 9:50 a.m.
In Response To: VRR  Challenge (Bob Koca)
Thanks for helping to explain. I am no so confident with Elo as a reliable indicator. On GG for instance, rating levels can easily be within a 200 elo range over a 3 year period for active players. If VRR was used what would the variance reduce to? 67 perhaps? Which would be a PR difference of 2.
I tend to think VRR winrate might be a superior way. You could normalise for average opponent strength by considering the average PR of all opponents played, convert into an Elo adjustment, using the standard formula to compute expected win rate vs a standard opponent (say 6.5 PR, the M3 threshold). Similarly, the results for any player could be normalised to a standard match length (say 11 points). This would seemingly resolve problems associated different player populations with varying strengths.
There is a practical issue of how the data from XG is used to populate the BMAB database. Sensibly the system makes use of XG player profiles which provide PR summaries for each session (BMAB/BMS Tournaments or Individual applications) but doesn't provide VRR or MWC/Cost results). How do you see this process working for VRR results? An operator painstakingly putting output parameters from each match?
This was why I set the challenge of proving a demonstration of the application of VRR to a 6 player round robin.

BGonline.org Forums is maintained by Stick with WebBBS 5.12.