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

BGonline.org Forums

Snowie-3, Snowie-4, Jellyfish, GNU and eXtreme rollouts all like 2pt with 64

Posted By: Steve Mellen
Date: Tuesday, 9 February 2010, at 5:41 a.m.

In Response To: Snowie-3, Snowie-4, Jellyfish, GNU and eXtreme rollouts all like 2pt with 64 (Paul Weaver)

Let us assume that these rollout results are 100% correct and that, with perfect play, making the 2-point is .005 better than the next-best play. I contend that it is impossible for the human mind, through casual observation, to notice such a small differential in actual play. This is likely true even if you randomize your plays in order to create a control group; it's certainly true if you always make the same play and thus have nothing to compare your results against!

Paul, when you write "I remember winning a lot of gammons in the games in which I opened with 64 and made the 2pt" I believe you are guilty of what the experts call confirmation bias. I do not think it is possible, even for an observant player like yourself over a period of 15 years, to notice that you are winning 1% more gammons with a certain play.

We all have our prejudices. Paul is surely right that some people are guilty of prejudice against making the 2pt because of what they were taught back in the day. I do not believe I am guilty of this, but I have a different prejudice; I believe it is silly to draw meaningful conclusions from tiny differences like .005 in opening-move rollouts, generated by bots which are clearly far short of having solved the game as a whole. Time will tell if my prejudice is irrational or not.

I am not suggesting that opening move rollouts are worthless, mind you. I am simply suggesting that sometimes we try to torture tiny differences into a conclusion when the reality is that some decisions remain "too close to call." If you have an imperfect bot that makes serious evaluation errors on a regular basis, you can do a trillion rollouts and you still won't be able to bake those evaluation errors out of the result with enough confidence to make differences at the third decimal place into anything meaningful.

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.