Thanks for the rollout, Neil.
Ten years ago, when I told people about 43S-61X being a strong play, I was met mostly by skepticism. At the time, the bots evaluated P as clearly best, while my rollouts showed X to be only a tiny underdog to P (and logically therefore hitting had to be right at certain scores). As bots got stronger, X's results pulled ahead a bit.
Data for 43S-61:
Snowie .. [X P9 H23] 31k
GnuBG .. [X H13 P17] 15k, d[X P13] 5k, g[H X9 P62] 5k
XG 3-ply [X P3 H9] 62k
XG 4-ply [X P10 H17] 12k (just posted)
To the right of the money data (GnuBG only) is DMP and GG data (where you see "d" and "g" preceding the brackets). At GS (not shown), I have only Snowie data (checker play not according to score): s[P X25 H32] 31k.
Nactation key: X = hit-and-split 24/18 6/5*, P = point 13/7 8/7, H = Hit-and-down 13/7 6/5*. As usual, the play that wins the rollout is listed first, followed by the next best play, etc., with its error size in thousandths of a point. The "k" number after the brackets is thousands of trials (rounded down).
For example, "[X P9 H23]" 31k means that X (hit-and-split) is best, P (point) is -.009, and H (hit-and-down) is -.023, after 31,000+ iterations.
After 43S-61X, how do you play the follow-up roll of 31? Diagram below.
| Money | 160 |
|---|---|
| |
| 43S-61X-31 | 165 |
Nack