BGonline.org Forums

XG and the Kauder paradox --- surprise!

Posted By: Phil Simborg
Date: Friday, 16 January 2015, at 12:21 p.m.

In Response To: XG and the Kauder paradox --- surprise! (Timothy Chow)

While I appreciate the intellectual quality of your post and this problem, from a practical standpoint would the Kauder Paradox come up in more than 1 in 10,000 games?

Do we really expect Xavier to amend his amazing program to take Kauder into account if I am correct and it hardly ever comes up? Also, when it does come up and XG fails to Beaver, it is very likely that it still got the double part of the cube action right and very likely that it had no affect on any checker play, so we are seeing an error only on the take decision. And how big an error would it be not to beaver?

Also, this only affects money games, so for the most serious backgammon that is analyzed the most carefully--match play/tournament play--this won't matter at all.

I have discussed many things with Xavier that XG can change to be either more accurate or more illustrative, and one of the criteria that Xavier uses (correctly in my opinion) is how many people will get benefit from the "improvement" and how important is the improvement compared to the time and effort required to make the improvement. In other words, he must prioritize in order to not make his program unnecessarily more "clunky" and complex so that it continues to be user-friendly, not take too long to do rollouts and analyses, and not be too complicated for the user to understand.

We use Variance Reduction, for example, knowing full well it is not quite as accurate, but it is accurate enough not to change a decision and saves tremendous time in rollouts.

Xavier's goal is NOT to produce the most completely accurate software in the world, but to produce a user-friendly software that provides highly accurate information in a reasonable period of time and for a reasonably-low cost. He has accomplished this in a fashion that most of us still can't believe he can deliver for only $60.00 (compared with Snowie for $400 and even when you compare it with GNUBG that is free when you consider XG's features).

So I do not take issue with your premise that XG will not get it right when a Kauder appears, but I don't believe that in any way diminishes the value of XG even if Xavier never incorporates Kauder into the rollouts.

Messages In This Thread

• XG and the Kauder paradox --- surprise!
  Timothy Chow -- Wednesday, 14 January 2015, at 10:53 p.m.
  • XG and the Kauder paradox --- surprise!
    ah_clem -- Thursday, 15 January 2015, at 7:12 p.m.
    • XG and the Kauder paradox --- surprise!
      neilkaz -- Thursday, 15 January 2015, at 8:37 p.m.
      • Simple solution if you don't like Jacoby
        Phil Simborg -- Friday, 16 January 2015, at 12:22 p.m.
    • XG and the Kauder paradox --- surprise!
      Timothy Chow -- Friday, 16 January 2015, at 12:24 a.m.
  • XG and the Kauder paradox --- surprise!
    Phil Simborg -- Friday, 16 January 2015, at 12:21 p.m.
    • XG and the Kauder paradox --- surprise!
      Bob Koca -- Friday, 16 January 2015, at 4:44 p.m.
  • XG and the Kauder--maybe only checker play rollouts?
    Phil Simborg -- Monday, 19 January 2015, at 2:15 p.m.
    • XG and the Kauder--maybe only checker play rollouts?
      Timothy Chow -- Monday, 19 January 2015, at 7:00 p.m.
  • XG and the Kauder paradox --- surprise!
    Casper Van der Tak -- Tuesday, 20 January 2015, at 12:23 a.m.
    • XG and the Kauder paradox --- surprise!
      eXtreme Gammon -- Tuesday, 20 January 2015, at 4:14 p.m.
      • XG and the Kauder paradox --- surprise!
        Casper Van der Tak -- Tuesday, 20 January 2015, at 10:54 p.m.
      • What is on the top of your list?
        niuniuch -- Thursday, 22 January 2015, at 12:05 p.m.