Re: [Bug-gnubg] Re: Naw..panic or teach me what I am doing wrong !!

[Joseph Heled]

but now it gives me strange results. The difference seems to be when 
using (or not using) the truncation. Can you try those 2

set rollout bearofftruncation exact off
set rollout bearofftruncation onesided off
set rollout quasirandom off

set analysis chequerplay rollout cubeful off
set analysis chequerplay rollout varredn off
set analysis chequerplay rollout quasirandom on

set rollout trials 144
set rollout seed 1

new game
set board 2LYBCDPOewMAAA
rollout



And

et rollout bearofftruncation exact on
set rollout bearofftruncation onesided on
set rollout quasirandom off

set analysis chequerplay rollout cubeful off
set analysis chequerplay rollout varredn off
set analysis chequerplay rollout quasirandom on

set rollout trials 144
set rollout seed 1

new game
set board 2LYBCDPOewMAAA
rollout


I get wildly different results. Perhaps the evaluation somewhere else is 
  wrong?

Joern Thyssen wrote:
> On Fri, Jan 10, 2003 at 10:06:29AM +1300, Joseph Heled wrote
>
> > I think I found the bug, but I would like an opinion because I can't 
> > believe such a blatant error can exist undetected for so long.
> >
> > I think wins are logged to the wrong player in the rollout code, when 
> > the game is over.
> >
> > line 747 in rollout.c
> >
> >      if ( iTurn & 1 ) InvertEvaluationR ( aarOutput[ ici ], pci );
> >
> > If I change it to
> >
> >      if ( ! (iTurn & 1) ) InvertEvaluationR ( aarOutput[ ici ], pci );
> >
> > I start to get exactly my numbers again.
>
>
> It think it's close enough: actually the bug appears to be that the
> evaluation just preceding that is evaluated for the wrong player on
> roll --- but your fix is equivalent, as we only end up there if the game
> is over, in which case it doesn't matter who's on roll.
>
> Jørn