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

BGonline.org Forums

XG cube analysis -- Methodology question

Posted By: Rick Janowski
Date: Wednesday, 5 December 2012, at 2:12 a.m.

In Response To: XG cube analysis -- Methodology question (Henrik Bukkjaer)

It is my understanding XG handles cube efficiency in a relatively simple but extremely effective manner. Firstly, it uses a single estimate of x (typically about 2/3, but may vary for races) for estimating cubeful equities at the 1-ply level. Any subsequent higher ply-evaluations maintain this cubeless to cubeful equity conversion. The effect of raising the ply-level is fortuitously equivalent to an iterative approach to rapidly improving/refining equity estimates. When the x-values (different for both players) are back-calculated at the higher play levels they appear to be relatively insensitive to the original estimate of x applied at the base-level. XG+ and XG++ evaluation are particularly effective in this regard. Examination of a wide range of race positions using the completely accurate Sconyers Database (all 15 vs 15 positions in the home board) shows the higher level evaluations to have only very occasional marginal errors, in almost all circumstances. Conventional rollouts using XG2 default settings produce cubeful equity estimates which are almost always within +/- 0.005 of the true equities derived from the Sconyers Database (usually with +/- 0.002).

Consequently, with regard to XG analysis and rollouts, the benefits gained by providing a more refined estimate of cube efficiency to be applied at the base-ply level, would not appear to be very significant.

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.