Re: [Bug-gnubg] Measuring performance levels

[Morten Wang]

* Morten Wang
> Rollout cubeless equity  +0,516
[snip]
> Full cubeful rollout with var.redn., with rotate rolls
> 21600 games, seed 11 using Mersenne Twister
> Play and cube: 0-ply cubeless
* Joseph Heled
> I think Doug insisted on Cubeless. All my rollouts show around a 3.7 gammon
> rate, while (if I read this correctly) this rollout gives 6.9
>
> So, is this really a cubeless rollout? What else can be wrong?

what I find odd here is that the cubeless equity is correct (it _was_
+0,516 for that rollout) and that the rollout settings are correct
except for the fact that it was a cubeless rollout (it is the only
full 21600, 0ply rollout I've run).

is gnubg reusing older cubeful rollout data?

I'm also wondering why there isn't a Copy-button in the rollout window
for easy transfer of the rollout data to the clipboard (there's only
OK, Stop & View statistics).  I tried to look for the rollout results
being available in another window, but couldn't find any (and the Hint
window looks confusing by now).

I've done one 1ply rollout by now and the result was as follows:

Win     (g)     (bg)    L(g)    L(bg)   Cubeless        Cubeful
0.7104  0,0246  0,0004  0,0779  0,0000  +0,3678         n/a
0,0007  0,0006  0,0001  0,0005  0,0048   0,0018         n/a

(settings: 21600 trials, 0 truncation, cubeless, var. red., rotate
 rolls, seed 11, 1ply, huge search space (8 cand, 0,160 tolerance),
 cubeless chequer eval and Jacoby rule off)


Morten!

-- 
"...a liquid which was almost, but not quite, entirely unlike tea."
  My opinions are not necessarily those of my provider,
  not necessarily mine, and probably not necessary.