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

BGonline.org Forums

Rollouts - Long

Posted By: Stick
Date: Thursday, 1 January 2009, at 11:36 p.m.

In Response To: Rollouts - Long (Rich Munitz)

I would also be just as agreeable that the different seeds are not the cause of the discrepancy

I have seen too much of this to agree. If I'm pointing my finger somewhere, that's where it's being pointed at the moment. It could be something completely different but that's my best guess.

If we look at the plays made with each setting for the 21 dice rolls in reply to the two opening 64 plays, does GNU play the same in all cases?

Supremo is a larger move filter, not a higher ply, so unless you think a move is being missed in the upcoming sequence by looking at 'only' 12 moves within' .160 v. 16 moves within' .320 ... not likely at all.

Also, this isn't a case of David's rollout in comparison to mine in this one case, I've seen it elsewhere where the rollout settings were exactly the same the only tangible difference being the seed number. (or perhaps the version of GNU but the same neural net)

I've always wondered whether the "adjustments" introduce bias into the results due to misunderstandings of the equities

If you want a deeper look into variance reduction, check out the links in this post: http://www.bgonline.org/forums/webbbs_config.pl?read=31549

Stick

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.