[Bug-gnubg] QUERY: Variance Reduction when rolling out Initial Position

[Ian Shaw]

There has been some discussion on GoL recently about rollout as initial 
position (RAIP).

People suspect that the VR doesn't take into account the RAIP setting, and 
reduces the variance assuming that doubles can be rolled.

Can someone confirm whether this is the case, please. 

If so, what will it mean for large rollouts I have already done? Will the 
rollouts be invalid or just the standard errors be too small/large. (Too large 
I would guess, since there would be more variance if doubles were allowed.)

See Zorba's post, below
--Ian

NEW: Re: Possible GNU Rollout Bug (Variance Reduction?)
Robert-Jan Veldhuizen (Zorba) -- Wednesday, 25 June 2003, at 5:46 p.m.

Well I did a few experiments, and it seems like you're right that it is the 
variance reduction causing problems here. The rollout trials won't roll any 
doubles on the first roll with the "RaIP" option enabled, but the variance 
reduction doesn't seem to know about this and thinks you're unlucky not rolling 
any doubles on the first roll. So, it will "correct" the equity and the end 
result is that it doesn't seem to make much difference whether you can roll a 
double on your first turn.