I just want to test how can I do, playing EXCLUSIVELY backgames in 1-pt matches. In this thread, I will post the results of the matches, the ratings of the opponents and GNUBG's overall analysis of the matches.
Please be constructive and don't make this thread a place for personal 'score settlements'.
Regards
1st match report:
Opponent's rating : 1546
Anchors secured : 1-pt & 2-pt
Match result : 1-0
Current rating : 1510
Change : + 10
2nd match report:
Opponent's rating : 1577
Anchors secured : 2-pt & 3-pt & 5-pt
Match result : 2-0
Current rating : 1520
Change : + 10
3rd-match report :
Opponent's rating : 1395
Anchors secured : 2-pt & 3-pt
Match result : 1-0
Current rating : 1529
Change : + 9
4th-d-match report :
Opponent's rating : 1502
Anchors secured : 1-pt & 2-pt & 4-pt
Match result : 1-0
Current rating : 1539
Change : + 10
5th-match report :
Opponent's rating : 1605
Anchors secured : 3-pt & 5-pt
Match result : 0-3
Current rating : 1530
Change : - 9
6th-match report :
Opponent's rating : 2051
Anchors secured : 3-pt & 4-pt
Match result : 1-0
Current rating : 1542
Change : + 12
7th-match report :
Opponent's rating : 1526
Anchors secured : 1-pt & 3-pt & 5-pt
Match result : 0-2
Current rating : 1532
Change : - 10
PS: This guy was amongst the luckiest noobs I've ever seen. Just see the match discourse. If we are to trust GNUBG's analysis in OTHER-THAN-MAASIVE-BACKGAME-SITUATIONS, I was a 97% favorite in the last stages, before he rolled some extraordinary lucky rolls.
8th-match report :
Opponent's rating : 1500
Anchors secured : 1-pt & 2-pt & 5-pt
Match result : 1-0
Current rating : 1542
Change : + 10
9th-match report :
Opponent's rating : 1405
Anchors secured : 1-pt & 5-pt
Match result : 0-2
Current rating : 1531
Change : - 11
10th-match report :
Opponent's rating : 1699
Anchors secured : 3-pt
Match result : 0-1
Current rating : 1523
Change : - 8
11th-match report :
Opponent's rating : 1520
Anchors secured : 1-pt & 2-pt & 4-pt
Match result : 1-0
Current rating : 1532
Change : + 9
12th-match report :
Opponent's rating : 2007
Anchors secured : None.
Match result : 3-0
Current rating : 1545
Change : + 13
13th-match report :
Opponent's rating : 1564
Anchors secured : 1-pt & 3-pt & 5-pt
Match result : 0-3
Current rating : 1535
Change : - 10
14th-match report :
Opponent's rating : 1490
Anchors secured : 1-pt & 4-pt
Match result : 1-0
Current rating : 1545
Change : + 10
15th-match report :
Opponent's rating : 1691
Anchors secured : 3-pt & 4-pt & 5-pt
Match result : 1-0
Current rating : 1555
Change : + 10
PS: Match log distorted.
16th-match report :
Opponent's rating : 1502
Anchors secured : 3-pt
Match result : 0-2
Current rating : 1545
Change : - 10
a thought has occured... IF gnuBG misanalyses backgames THEN using it for the analysis won't work will it? How about using a different neural net which has been optimised for 1 pointers? Zorb will know how to switch the nets GnuBG uses - perhaps try and get your hands on the one eg MonteCarlo uses?
Any views yet from the experts?
Only that Monte Carlo is considerably weaker than gnu, snowie and bgblitz and it doesnt do analysis anyway.
My point was not on its' strength of play but that its' net will see far more 1 pointers. I know it doesn't do analysis - hence my thought on switching the neural net embedded within GnuBG.
On the related issue that GnuBG et al will outplay MonteCarlo that doesn't tell us much if MonteCarlo is selecting its' moves with some noise in the selection process does it?