| |
|---|---|
| Blue on roll after White's opening 43U | |
It turns out that Snowie (even though checkers aren't played according to score) did a pretty decent job on these. Below your list is pasted, this time with Snowie data added after the comma in each row, for reference. All rollouts are 5k (5184 trials) unless otherwise marked:
43U-44 d[K P6], d[K P3]
43U-41 d[K X15], d[K X21]
43U-61 d[X H3 P22], d[H X1 P9] 20k
43U-62 d[H h11 X15], d[H X11 h16] 10k
43U-43 d[X K8 x22], d[X K11 x24] 31k 5kx
43U-54 d[H S3], d[H=S] 20k
43U-64 d[X H1 P6 $9], d[H X4 P4 $4]
43U-65 d[S R11], d[S R9] 20k
For purposes of AtS comparison, below are money rollouts of the same positions. All are Snowie; two (43U-62 and 43U-43) have XG data averaged in.
43U-44 [K X50] 5k
43U-41 [K X58] 5k
43U-61 [P H10 X22] 20k
43U-62 [H h35 X41] 41k 10kX
43U-43 [X K29 x37] 62k 20kx
43U-54 [S H2] 20k
43U-64 [P $8 H16 X33] 5k
43U-65 [S R22] 25k
Key: H = Hit, h = hit on lower point, X = Hit-and-split, x = hit-on-lower-point-and-split, K = Kill (hit twice), P = Point, R = Run, S = Split (and down), $ = Slot. Error sizes are in thousandths. Before brackets, d = dmp; else assume money.For example, "d[H S3]" means that at dmp, H is best and S is -.003.
Stick says he remembers H beating P and S in the 43U-54 DMP rollout he can't find. That sounds right: For reference, Snowie is d[H S12 P14] 31k. For money, agg (bot aggregate) is 43U-53 [H P11 S20].
Based on the (fairly extensive) money data and XG-5k-only DMP data we have so far, then, positions for which the best DMP plays differs from the best money play are:
43U-61: [P H10 X22], but d[X H3 P22]
43U-54: [S H2], but d[H S3]
43U-64: [P $8 H16 X33], but d[X H1 P6 $9].
Nack