|
BGonline.org Forums
Thanks! +1
Posted By: Christian Plenz In Response To: Thanks! +1 (Maik Stiebler)
Date: Tuesday, 8 May 2012, at 2:42 a.m.
The PR is not direcly displayed in the export. BTW this might be a part of improving Extremegammon. The statistics and player profiles are great but sometimes you have to do some more steps with not only Extremegammon but also Excel/Openoffice. You get 'Eq per Decision'. Roughly this number times 500 should result in the PR. I have weighted it by the number of decisions per game/Match and for the 10 real 1-Point-Matches the calculated PR matches with the displayed PR if I choose Match Lenghts: 1 Pt in the results tab.
Furthermore I think that the factor of 1.5 is already included in the 'Eq per Decision'. For instance I have compared the same game as a 1-Point-Match and with copy&paste as DMP at 6-6 to 7. When I look at the 7-Point-Match game per game in the Match Summary the PR for the DMP as well as the Level of Play for this last game is lower by a factor of 1.5 on comparison the the 1-Point-Match. The exported Game list for the 550 matches looks like the same (the 'Eq per Decision' in DMP game is always factored by 1.5 when it goes from displayed Match Summaray to the Player Profile whereas the real 1-Point-Match remains the same)
Maik writes: 'The interesting question is: How does your expected win rate of 54.64% of one-pointers against your typical opponent compare to your performance against the same typical opponent in longer matches? '
This is something I have to sleep over. So far we have 54.64% in 57 lets say 1-Point-Comparisons and 58.04% (thats taken from the 'Should have won') in 550 6.5-Point-Matches.
The PR difference for all matches is 3.64.
3.64 * 33 = 120 ELO
120 ELO in a 6-Point-Match -> 58.4% 120 ELO in a 7-Ponit-Match -> 59%
So the ELO formular overestimates me in the longer matches or underestimates me in the 1-Point-Matches or both :) BTW Opponents ELO taken from Gridgammon is 6 Points higher in the 1-Point-Matches.
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.