|
BGonline.org Forums
Average # of Blunders per Match (Giant)
Posted By: Daniel Murphy In Response To: Average # of Blunders per Match (Giant) (Bill Riles)
Date: Tuesday, 21 February 2012, at 12:22 a.m.
Well, Stick asked for opinions about a Giant 32's play against a random human opponent, so there would be various human factors to consider in the Giant's play. But having said that, as an experiment I wanted to test (or illustrate) what I think everyone knows: that it's possible to blunder fairly often and still play very well overall.
I had Gnubg 0-ply play a 9 point match but made Player 1 blunder on every opening response (by coincidence Player 0 won all 8 opening rolls.) Each blunder was as close to -0.080 as I could get. Then I had Gnubg 0-ply finished each game. Then I analyzed the match with 2-ply.
So, Player 1 made 1 blunder per game. Player 1's match statistics:
Gnubg ER: -7.35 Snowie-equivalent ER: -3.01 XG PR should, I think, tend to be about half the Gnubg ER, so about -3.7.
So for what it's worth, it's possible for a bot to blunder once in every game in a 8-game, 9 point match, play all other moves as Gnubg 0-ply, and still have ER under 4.0.
Player 1's moves Total moves: 191 Unforced moves: 148 Doubtful moves: 8 Bad moves: 4 Very bad moves: 6 Cube errors: 2
Player 1's errors -0.151 (52 13/6 in response to 21$ -0.111 (23 13/11 8/5 in response to 61P) -0.094 (56 13/8 13/7 in response to 52D) -0.087 (51 24/18 in response to 53P) -0.087 (21 24/22 6/5 in response to 51S) -0.080 (63 13/10 13/7 in response to 21$) -0-080 (22 13/11 8/4 6/4 in response to 21S) -0.079 (11 24/23 6/5(3) in response to 53P) -0.056 -0.051 (wrong double) -0.048 -0.037 -0.034 -0.026 (wrong double) -0.023 -0.023 -0.022 -0.022 -0.022 -0.014
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.