On 7/2/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
When GNU-bg reports a cube rollout result it shows (I believe) the cubeless equity as calculated from the s/g/bg report of the no-double branch. Wouldn't it be more accurate to instead use the breakdowns from the take branch, since (presumably) the number of dicerolls in that path is typically larger (and thus less susceptible to pure evaluation)?
It sounds reasonable, at least for money games. I suggest a couple of (10) test cases to see if at makes any difference. But, in general I don't look much at the cubeless equties, nor at the s/g/bg breakdown, since I often find that they make little sense, except where two very different moves result in the same cubefull equity. Christian. _______________________________________________ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg