[Bug-gnubg] Fwd: Long GNU rollout results

[Øystein Johansen]

Hi,

I'm forwarding this posting to GammOnLine by Neil Kazaross. Can someone 
answer?


---8<---8<---

Overnight I did 6480 games 1ply/cubeful with GNU which will play 
according to score. It shows that Douglas is correct that this just 
isn't good enough to cube, even at this score.

Anyhow..I remain confused about a couple of things with GNU rollouts.

1) Can someone, in detail. please explain how the numbers in the first 
line (cube analysis) are obtained.

2) Why are there always very slight differences in the rollout details 
between centered 1 cube and opp owns 2 cube ? This makes absolutely no 
sence at all. In a live cube position it cannot be possible nor accurate 
to measure rollout details as presented below since over a series of 
games, the cube will be at different levels. I'd like to be able to 
trust GNu rollouts 100% and be 100% sure that it is playing and handling 
the cube with out bugs in match situations but when I see the data 
presented in a way that makes no sence, I cannot have the confidence I'd 
like to have.

My appologies to the GNU programmers, but I simply want to have 100% 
trust in GNU.

Cube analysis Rollout cubeless equity +0.362 (Money: +0.343)

1. No double +0.490
2. Double, pass +1.000 ( +0.510)
3. Double, take +0.470 ( -0.021)
Proper cube action: No double, take (3.9%)

Rollout details:
Centered 1-cube:
0.599 0.278 0.010 - 0.401 0.130 0.014 CL +0.362 CF +0.490
[0.001 0.001 0.001 - 0.001 0.001 0.001 CL 0.003 CF +0.007]
Player gnubg owns 2-cube:
0.597 0.285 0.010 - 0.403 0.135 0.022 CL +0.796 CF +0.470
[0.001 0.002 0.001 - 0.001 0.001 0.001 CL 0.007 CF +0.009]
Full cubeful rollout with var.redn. 6480 games, seed 1195169792 using 
quasi-random dice and Mersenne Twister Play and cube: 1-ply cubeful