[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Which play is better based upon rollouts

Posted By: David Rockwell
Date: Tuesday, 27 April 2010, at 1:17 p.m.

In Response To: Which play is better based upon rollouts (Michael Depreli)

Assume that the opening roll is 61 which is played by making the bar point. What are the odds that responder will play a backgame? What are the odds that the opener will play a backgame? I don't have data on this and I am not going to waste time on this. But, I would be shocked if backgames aren't divided close to 50-50 between the two players. What about a 31P opening? This would have more skew in it. But, I would still expect the opener to play 40% or more of the backgames that develop. After an opening 64S, can we even guess correctly which side is more likely to play a backgame? If both players are playing backgames, then both sides of the rollout are generating errors from backgames and they will tend to be offsetting.

Note too that everyone is assuming that the bot is playing one side of the backgame rollout perfectly while it butchers the other side. This won't happen either. It will make mistakes for both sides of the backgame which will also tend to be offsetting.

Now think about discounting. How quickly do backgames develop? An error 10 moves into the game that is .02 affects the final rollout result by .00000002 if I have placed my decimal point correctly. (I will disclose that I am not being careful with my decimals since the point is obvious anyway.) You need 5,000 of these errors, all in one direction, before this error would contribute .0001 worth of error to the rollout result. If the backgame develops in 20 moves rather than 10 moves, the error would be .0000000000002 and you would need 5,000,000,000 of these errors, all in one direction, to contribute .0001 worth of distortion to the rollout result. So, even when errors are not offsetting, they would have to be quite numerous indeed to have a meaningful effect on the final result.

Messages In This Thread

 

Post Response
Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:
 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

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