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

BGonline.org Forums

Does XG's definition of PR annoy you?

Posted By: Timothy Chow
Date: Tuesday, 9 July 2024, at 6:44 p.m.

In Response To: Does XG's definition of PR annoy you? (Bob Koca)

Bob Koca wrote:

It makes it sound like there is a reason it could be very close but not literally the same but I don't see why.

I'm taking the definition of "PR" to be "what the eXtremeGammon program outputs as the PR" as opposed to "what the eXtremeGammon documentation claims the definition of PR is." In particular, your PR for a match depends on the neural net that is doing the calculations.

Even if we went with the documentation rather than the code, I'm not sure the documentation completely specifies how to calculate the PR. For example, when determining whether or not a particular move is an error, how exactly does the move filtering work? Suppose we have a position where the roll is 11, and it can be played hundreds of different ways. Do we know exactly how XG picks its 32 top candidates? If the 32nd and 33rd plays are exactly tied in equity according to the neural net, how is the tie broken?

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.