|
BGonline.org Forums
Backgammon, Poker, and PRs
Posted By: Florin Popa In Response To: Backgammon, Poker, and PRs (Robert Wachtel)
Date: Sunday, 28 February 2016, at 12:39 a.m.
"PR rankings, as a number of commentators have observed, may fall short of perfection. A common complaint, heard at all levels of achievement, is: “I play meta-backgammon. I know the ‘right’ plays, the ‘computer moves,’ but I deliberately deviate from them to give my opponents more problems. I bluff-double, I create complicated positions and pass clear takes when the position is too easy for my opponent to play. The bot cannot appreciate my strategy, and penalizes me for correctly taking advantage of my opponent’s weaknesses.”
In my opinion, claims of this type are usually highly exaggerated. In general their proponent grossly overestimates the return on investment (both in magnitude and in likelihood) for the equity sacrifices he (deliberately?) makes – and often his sacrifices are not deliberate! He is guessing – and blundering "
I agree with you that most of the times are exageration but two things are clear for me 1) if you create more complex position your PR and opponent's one will suffer. 2) XG sugestions are based on the fact both players are PR0. I remeber when strong computer programs came on the market players claimed they had winning position because computer indicated big advantage. But they lost those matches not being able to find the right continuation. Ok, in bg is not such a big difference in strengthness as is chess between a software and an average player but still some difference exists between a PR0 and a PR7. So I think the human factor in bg plays a bigger role than many people think but we cannot measure it. I am not 100% sure XG is favourite to perform better than best human players in many tournaments, it could take sometimes bad decisions against weak player like redoubling to 4 in a race, keeping the cube would be a better decision considering 7 or more points difference in PR.
|
BGonline.org Forums is maintained by Stick with WebBBS 5.12.