| |
|---|---|
| 52S-64 | |
For gammon save, the shortcut is to compare wins minus gammon losses (or losses plus gammon losses, works the same way). Most of the time, it's a pretty good indicator. However, the 52S-64 at GS position shows that you can't rely on this indicator.
As I recall, some older version of Snowie could roll out positions with checkers played according to score. Snowie-4 cannot. In the pre-GnuBG days, the best we could do is to roll out the score for money and extrapolate (actually, Snowie does the math for us), sometimes leading to erroneous conclusions.
The Snowie (money) rollout for 52S-64, H (Hit, 13/3*) has 46.9% wins and 15.2% gammon losses, netting 31.7. By contrast, R has 47.0% wins and 14.8% gammon losses, netting 32.2, for a difference between the two nets of 0.5. To translate to nacbrac margins, double and bounce the decimal point to the right, yielding [R H10]. In other words, the bot is telling us that if the checkers are played as if for money, then at gammon save R beats H by .010.
The Gnu money rollout produces 46.96 - 15.35 = 31.61 for H, and 47.02 - 14.5 = 32.52 for R, for a difference of 0.91, double and bounce (and round) comes to [R H18].
The XG (4-ply) money rollout produces 47.09 - 15.7 = 31.39, and 47.02 - 14.85 = 32.17, difference of 0.78, double and bounce, comes to [R H16].
Recapping, the win-minus-gammon-loss method applied to money results for Snowie, Gnu and XG would have one predict that gammon save results should be [R H10], [R H18] and [R H16], respectively. But the AtS rollouts belie that method by producing results of [H R4] and [H R2] -- Gnu and XG respectively.
Nack