I had missed this thread entirely because I forgot the joke about "unreal backgammon." (Mostly I read only posts that have to do with early game rollouts or analysis.) Thanks for posting the Nactation sequences in the titles of the rollout posts -- drew me like a magnet. Great set of positions!
For three of the five positions, I found Snowie rollouts, though fewer trials and no live cube. I've included those to the right of your XG results summarized below:
42P-61P-42 [S Z2] 20k, Snowie [Z S2] 10k
42P-31P-42 [Z R33 D35] 2k
42P-42P-42 [S Z3] 20k, Snowie [S Z3 D11] 5k
42P-53P-42 [D S15] 20k
42P-64P-42 [D Z19 S27] 10k, Snowie [D Z24 S40] 5k
Key: S = Split (24/20 13/11), Z = reverse split (24/22 13/9), D = Down (13/11 13/9), R = Run (24/18). Error sizes are in thousandths. After the brackets, the number of trials (in thousands, rounded down) is listed. For example, "[D Z19 S27] 10k" means that D is best, Z is -.019 and S is -.027 after 10000+ trials.
John and Chase seem to have nailed all the right answers (and good analyses accompanying), though in the case of the first and third position (for which they chose S) I think it's more meaningful to say that S and Z are tied.
I was a bit surprised that for the 42P position you didn't roll out D (which Snowie's rollout has at only -.011).
For reference, the five positions are diagrammed below, with (aggregate) rollout results encaptioned.
| |
|---|---|
| 42P-61P-42 [S Z1] | |
|
| ||||||||||||||||
|
| ||||||||||||||||
For the four inside-point positions, note how the edge by which split beats down changes (crossing over between 42P and 53P): .035, .011, -.015, -.031.
Nack