Beating XGR+ in a long unlimited session?
Posted By: Daniel Murphy In Response To: Beating XGR+ in a long unlimited session? (Stick)
Date: Wednesday, 12 September 2012, at 7:39 p.m.
In Response To: Beating XGR+ in a long unlimited session? (Stick)
From a 2009 BGO discussion of Matt "I will play money games against GNU, Snowie, or XG until my fingers cramp, and after that I can call out my moves" Cohn-Geier's challenge in the 2+2 forum to play Snowie for money if Snowie found some backers:
If the bot is playing with a 0.0 ER, what will MCG's ER have to be in order for him to be the favorite?
I don't know what it has to be -- that is, how low it could be with Matt still winning. But Matt doesn't need a low ER to beat a bot. In three short sessions I played in October against Gnubg with the intention of exploiting its cube and strategy weaknesses, my results were:
Games Points Won Snowie-equivalent error rate 9 +10 -18.55 10 +14 -25.63 29 +30 -24.30
I'm not keeping any secrets. I've discussed how to beat Gnubg by exploiting its poor play against and evaluations of massive backgames and rolling prime positions many times before. These are, I believe, Gnubg's most serious and exploitable weaknesses, They are not its only weaknesses, which include its misevaluations of holding games, and of games in which one side has many checkers off, but knowing of those faults won't get you mega-wins.
And from 2008:
My recollection is that a year or more ago, perhaps early Summer 2007, I played money games vs. Gnubg, seeking to play a massive backgame nearly every time, and soon racked up a huge point advantage. My success was due partly to Gnubg's willingness to give me happy-beaver cubes just because I had five or eight or eleven men back, and mostly because of Gnubg's later gross misevaluations of cube action in post-backgame post-hit snakelike positions.
After I posted some positions and session results at Gammon University, Casper and one or two other Readers posted that they were similarly able to beat the bot this way. I don't remember when exactly that was, but in August '07 Stick quoted me as having written: "I'd back myself against Gnubg in a money session, and I think most any expert-whatever player should be able to similarly exploit Gnubg's weakness with cube decisions in massive backgames and snake positions."
Messages In This Thread
BGonline.org Forums is maintained by Stick with WebBBS 5.12.