Re: [Bug-gnubg] Rollout bugs (j.s.d. options)

[Jim Segrave]

On Thu 31 Jul 2003 (16:42 +0000), Joern Thyssen wrote:
> On Thu, Jul 31, 2003 at 04:13:55PM +0200, Jim Segrave wrote
> > 
> > I really can't reporduce this and nothing that could affect it has
> > been changed since well before 27 July. Very odd
> 
> I can reproduce it here.
> 
> set rollout cubeful on
> set rollout varredn on
> set rollout quasirandom on
> set rollout initial off
> set rollout truncation enable on
> set rollout truncation plies 6
> set rollout bearofftruncation exact on
> set rollout bearofftruncation onesided on
> set rollout later enable off
> set rollout later plies 5
> set rollout trials 1296
> set rollout cube-equal-chequer on
> set rollout players-are-same on
> set rollout truncate-equal-player0 on
> set rollout limit enable off
> set rollout limit minimumgames 144
> set rollout limit maxerror 0.1000
> set rollout jsd stop on
> set rollout jsd move off
> set rollout jsd minimumgames 36
> set rollout jsd limit 1.9600
> set matchid cIkFAAAAAAAA
> set board 4HPwATDgc/ABMA
...

Sorry, I misunderstood what was being reported. I thought that the
reported problem was that the rollout was doing one more trial than
requested (so for the above it would be doing 1297 trials).

There is a bug - the test for minimum trials is against the loop
counter which starts at 0 and is only incremented after the j.s.d. and
std. error limit processing has been done. I'll commit a fix today.

-- 
Jim Segrave           address@hidden