| |
BGonline.org Forums
XGR++ RO and 1 ply RO..etc
Posted By: Henrik Bukkjaer In Response To: XGR++ RO and 1 ply RO..etc (Henrik Bukkjaer)
Date: Thursday, 15 December 2011, at 1:17 p.m.
PS:
I honestly think that this area is where the bots could be HUGELY improved. They play so close to perfect now, that potential improvements in optimal game-play is very minor stuff. But think of Sticks post, where he states that he only think about 10 players in the world qualifies for the top-32!! :-) Well, what it says is, that there are VERY FEW actual match-ups played, where the players are in the same ball-park as optimal bot-play. For all those other match-ups, the resulting analysis we get, is not optimal. So in order to improve on the win-rate, rather than the PR, we need bots and tools to help us in all those non-perfect situations.
Personally, I'm rated around no. 25 in DBgF currently. I have no chance of beating XG in the long run, playing 17 point matches (my error-rate is averaging just above 4, with the occasional sub 3.5, but also the occasional 6+). However, I'm fairly confident, that I would outperform XG if we sat down and played say 100 matches vs. people ranked below no. 300 on our rating list. That's people playing with error rates starting from 7 and up.
Why? Because XG doesn't adapt to the opponent at all, it's not aware that it's favorite to win, thus it uses a hopeless match equity table to determine it's cube actions, and so on...
I know that bet is virtually impossible to carry out in real life, but would there be any takers?
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.