BGonline.org Forums

53P-21 at Match Scores

Posted By: Nack Ballard
Date: Thursday, 22 July 2010, at 11:52 p.m.

In Response To: 53P-21 at Match Scores (Stick)

I know GNU has a slotting preference in the opening game but I think it gets this one right, I slot at gammon save here.

By "right," I assume you mean on what side of zero value is -- which play is best. Okay, but I'm also interested in how far off of value that bots might be in their rollouts, and in making sense of related results.

Consider opening 21. For money, Gnu is [$ S19] 92k, down to [$ S14] after adjusting for Gnu's misreplies (David's research), whereas Snowie is [$ S8] 80k, XG3 is [$ S5] 82k and XG4 is [$ S3] 57k. I believe the difference (of .006, .009 or .011, plus the extra .005 in misreply) is an example of Gnu's (seemingly typical) rollout bias for slotting plays.

At GS (still opening 21), Gnu is s[$ S2] 61k, and XG is s[S $2] 66k. If we assume XG is "right" and Gnu is "wrong" (and assume variance is quashed), it's hard to say if the .004 difference is explained again by (a) Gnu's misreplies or by (b) a carryover in its money bias. Though it's only a mystery in passing at this point, between (a) and (b) I would have expected the difference between Gnu's and XG's GS results to be greater than .004 (which would translate to roughly .006 if this were on the order of a money comparison), but okay maybe Gnu plays better at GS than it does for money.

Now consider 53P-21. For money, Gnu is [S $2] 46k (with no research yet done for possible third-roll-error ajustment), Snowie is [S $11] 20k, XG3 is [S $11] 77k, and XG4 is [S $7] 22k so far. Compared to the other bots, Gnu seems to have a slot bias of around .008.

At GS (still 53P-21), Gnu is s[$ 15.3] 15k. While I am aware that when Opp has a deep point it is more reasonable to avoid splitting at gammon save, there is still the matter of examining the size of the difference and the direction in margin. According to Gnu's results, for opening 21, a player should be less slotty by .017 (or at least .012) at GS than for money. By contrast, for 53P-21, one should be more slotty by .017+ at GS than for money. Those adjustments in AtS checker play aren't necessarily inconsistent, but regardless, I'd like to get a better handle on it.

Nack

Messages In This Thread

• 53p-21 RO early results
  Alexandre Le Siege -- Thursday, 22 July 2010, at 7:59 a.m.
  • 53P-21 at Match Scores
    Paul Weaver -- Thursday, 22 July 2010, at 2:34 p.m.
    • 53P-21 at Match Scores
      Leonardo Jerkovic -- Thursday, 22 July 2010, at 3:05 p.m.
    • 53P-21 at Match Scores
      Nack Ballard -- Thursday, 22 July 2010, at 9:16 p.m.
      • 53P-21 at Match Scores
        Stick -- Thursday, 22 July 2010, at 10:40 p.m.
        • 53P-21 at Match Scores
          Nack Ballard -- Thursday, 22 July 2010, at 11:52 p.m.
          • 53P-21 at Match Scores
            Leonardo Jerkovic -- Friday, 23 July 2010, at 12:00 a.m.
            • 53P-21 at Match Scores
              Nack Ballard -- Friday, 23 July 2010, at 12:07 a.m.
          • 53P-21 at Match Scores
            David Rockwell -- Friday, 23 July 2010, at 12:11 a.m.
      • 53P-21 at GS 4 ply RO started *NM*
        neilkaz -- Friday, 23 July 2010, at 6:35 p.m.
  • 53p-21 RO early results
    Nack Ballard -- Thursday, 22 July 2010, at 8:56 p.m.
  • Thx so much !!
    neilkaz -- Saturday, 24 July 2010, at 1:51 a.m.
    • Thx so much !!
      Alexandre Le Siege -- Saturday, 24 July 2010, at 2:16 a.m.