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

BGonline.org Forums

Please also check for GNU..my quick trucated RO setting

Posted By: Michael Depreli
Date: Tuesday, 15 September 2009, at 6:43 p.m.

In Response To: Please also check for GNU..my quick trucated RO setting (neilkaz)

In order for the new test not to be biased, I'd have to run the 324 0/0 @6 for every move/cube decison in the 500 game series right?

I can see you're trying to emulate XG Roller in GNU. (I've done the same).

Here are my initial observations:

If you run XG Roller or GNU equiv on problems where you know there is a largish error factor as in the Depreli positions it will generally quite quickly find the best play. However if you run this type of eval on positions where say 2 (or more) choices are quite close it can often shuffle the moves around.

Examples of opening replies:

21S 32 correct play 24/21 13/11

XG 3 & 4 ply play this correctly XGRoller choses 24/22 13/10 an error

32S 43 correct play 8/4* 13/10

XG 3 & 4 ply play this correctly XG Roller choses 8/4* 24/21 an error

So the question is do these small errors get outweighed by the large "obvious" errors XGRoller doesn't make by doing a truncated rollout?

I can't be sure but it's just some initial observations which hopefully will be cleared up when all the full rollouts have been done for the new series.

Michael

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.