| |
BGonline.org Forums
$$ 16-cube reship -- with some bot data
Posted By: Daniel Murphy In Response To: $$ 16-cube reship (Alejandro)
Date: Friday, 21 May 2010, at 1:35 a.m.
(1) the takepoint increases, the higher the cube value according to GNU?
No. To verify this, edit your position, change the cube value to 2 or 4 or 256, and reanalyze. The data in the analysis window will all be the same, regardless of cube level.
(2) how can it be profitable to take at 81%?
If you neither win nor lose any gammons, your money game takepoint is 25%, ignoring recube vig, which generally reduces your takepoint to about (but above) 22%. But you must adjust that takepoint for gammons won and lost. How do you do that? In money games the gammon price is 0.5. Estimate your gammon losses, subtract from that your gammon wins, multiply the result by 0.5, and add or subtract that result to your gammonless takepoint. If you lose more gammons than you win, adjust your takepoint upwards. If you win more gammons than you lose, adjust it downwards. In other words, for every 2% net gammons (those of one player less those of the other), adjust your takepoint up or down by 1%, in favor of the player who wins more gammons.
If all your wins were gammons and none of them were losses, your takepoint would be only 16.66% (on a 2-cube you'd lose 2 points 5 times and win 4 points once for a loss of 6 points, the same as dropping six times).
In your position, obviously not all of White's wins can be gammons, since some of them will be re-re-re-redoubled passes.
(3) Unfortunately your bot's evaluation is wrong. This is a redouble/pass.
Cubeless -- that is, if all games were played to completion without White ever redoubling -- White would win about 17.5% of all games, including about 11.5% gammons. Using those numbers (until some better ones come along), if White takes he loses about 0.535 points times the cube value per game -- 1.07 points on a 2-cube -- if he never redoubles. In your money game position, White can win more points in some games by doubling Blue in, and win more single games by doubling Blue out, but doubling Blue out probably reduces White's gammon wins slightly. However, White doesn't gain enough with cube vig over his cubeless chances to take Blue's cube in the problem position.
(4) With regard to using Gnubg, generally 3-ply is fine for checker evaluations (3-ply is better overall than 2-ply, but both make mistakes that the other does not). But, again generally, 2-ply is much better at evaluating cube decisions.
That said, in lopsided positions 3-ply evaluations can be more accurate. In this position 3-ply is more accurate, but both 3-ply and 2-ply are wrong -- 3-ply says redouble/take and 2-ply says no double.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.