BGonline.org Forums

31P-32, data

Posted By: Nack Ballard
Date: Wednesday, 18 August 2010, at 10:07 p.m.

In Response To: 31P-32 4 ply RO (neilkaz)

31P-32

Thanks for the RO, Neil. Data for 31P-32 (money, dmp, save, go) is

Snowie .. [Z S6 U17 D40 T67] 31k 5kDT
GnuBG .. [Z S8.5 U20 D36] 46k 15kU 7kD
XG 3-ply [Z S22 U33 D50 T57] 10k 5kUDT
XG 4-ply [Z S7 U21] 38k 10kU

GnuBG .. d[S=U Z1] 10k
XG 3-ply d[S Z1 U1] 46k

GnuBG .. s[Z S5 U22] 15k 5kU

GnuBG .. g[D Z5 S13] 5k

Key: Z = reverse split (24/22 13/10), S = Split (24/21 13/11), U = Up (24/22 24/21), D = Down (13/11 13/10), T = sTack/Tower (13/8). Error sizes are in thousandths.

Before the brackets, d = dmp, s = save, g = go; else assume money. Number of trials is in thousands (k), rounded down. For example, "g[D Z5 S13] 5k" means that at (gammon) go, D is best, Z is -.005 and S is -.013, after 5000+ trials.

For money, Snowie, GnuBG and XG 4-ply closely agree (.006, .0085 and .007, respectively) on the Z/S margin, while XG 3-ply so far makes it .022 in this 10k rollout. [XG (3-ply) essentially claims that there is a 95% chance that margin will end up between .011 and .033 if rolled out a billion trials.]

Nack

Messages In This Thread

• 31P-32 4 ply RO
  neilkaz -- Wednesday, 18 August 2010, at 3:03 p.m.
  • 31P-32, data
    Nack Ballard -- Wednesday, 18 August 2010, at 10:07 p.m.