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

BGonline.org Forums

52S-C.-2-2 5184 XG Rollout Result Posted

Posted By: Timothy Chow
Date: Monday, 15 September 2014, at 1:25 a.m.

In Response To: 52S-C.-2-2 5184 XG Rollout Result Posted (Dmitriy Obukhov)

Dmitriy wrote:

One thing to notice in Paul's rollout, which is also present in mine, that cubeless numbers after double and no double are not the same. Since bot will never lose market, this seems strange to me. Can anyone explain me why two equities in the rollout – this is also true in my rollout – didn't converge?

After doubling, the bot will play the checkers as if it's DMP and will report as "cubeless" numbers whatever wins/gammons/backgammons result.

After ND, it's more complicated. I think you're right that if the bot were always playing the checkers the same way it would at DMP and if it never passed any of its own doubles, then you would get the same "cubeless" numbers coming out. (In particular I don't think that variance reduction affects the "cubeless" numbers. But this might be worth verifying, with a rollout with variance reduction turned off.) So there are two possibilities: (1) some of its checker plays when the cube is still not turned differ from its checker plays when the cube is turned, and/or (2) sometimes it loses its market.

Possibility (1) seems more likely, but I don't see why possibility (2) can't happen. It won't happen with perfect play, of course, but the bot doesn't play perfectly.

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.