
BGonline.org Forums
What is XG Thinking here?
Posted By: Tom Keith
Date: Sunday, 14 September 2014, at 1:08 p.m.
In Response To: What is XG Thinking here? (Dmitriy Obukhov)
This just may be an artifact of XG's conversion of the breakdown of wins, losses, and gammons to match equity. For example, suppose X wins a single game 35% of the time, wins a gammon 10% of the time, loses a single game 40% of the time, and loses a gammon 15% of the time. When the cube is at 2, it is easy to convert this to match equity:
Event Probability Match equity Product X wins 2 .450 +1.000 +0.450 X loses 2 .550 −1.000 −0.550 Total 1.000 −0.100 When the cube is at 1, you know and I know that the bot should still use the same formula because (assuming perfect play) the cube will end up on 2 anyway. But a bot might not know that; a bot might naively assume the cube stays at 1. In that case, the match equity calculation breaks down like this:
Event Probability Match equity Product X wins 2 .10 +1.000 +0.100 X wins 1 .35 +0.685 +0.234 X loses 1 .40 −0.685 −0.274 X loses 2 .15 −1.000 −0.150 Total 1.00 −0.090 So the choice the bot faces is "double", which has an equity of −0.100, or "no double", which has an equity of −0.090. By this calculation, "no double" is better by 0.010.

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