[Bug-gnubg] RE: Bug-gnubg Digest, Vol 57, Issue 11

[bob koca]

> From: "Christian Anthon" <address@hidden>
> 
> Hi Bob,
> 
> the rollout is truncated in two bearoff situations
> 
> a) when cubeful only for moneygame and only when the positon is in gnubg_ts
> b) when cubeles and the position is in either gnubg_os or gnubg_ts. If the
> positon is in gnubg_ts the actual lookup is done there.
> 
 
 
       Thanks for that info. If I was doing a DMP rollout and wanted to use the database
I would not have realized that I should have turned cubeful off. 
 

> 
> So for expert/expert you get the result of a real rollout in the first three
> columns and the result of a lookup in gnubg_ts in the last two columns in
> general. The data in gnubg_ts has two byte precision and therefore shouldn't
> be trusted beyond an accuracy of 0.0001.
 
         
     Thanks, I understand the small discrepancy for expert vs. expert now. 


> As for the remaining data I don't know and I don't care too much. I can see
> how the noise might enter the winning chances: if noise is added to the
> evaluation of game over positions that is if 1.00 is sometimes (for example)
> 0.91.
> 
> Personally I think that adding noise is a bit of an abomination, which has
> only small justification in human vs. computer play, and no justification
> whatsoever when doing rollouts. The data you obtain might be right, but I
> fail to see what they can be used for even when they are right. Setting GNU
> backgammon up to play against itself with different settings for the two
> players makes more sense to me.
 
 
         I agree that for how most use rollouts that there is no reason to add noise. For
my purpose though I want it. I am trying a test assumptions in the ELO rating system
and need to know win% for expert vs advanced, advanced vs. intermediate, and 
expert vs. intermediate.  
 
     I am not sure I understand your hypothesis for the noise making a difference when
the game has only forced moves like my example.  I had thought that the noise was only 
added during the choice of moves. Is noise added to the evaluation of game over positions also?
Possible I guess. No reason to do so though. Also  if that is the case wouldn't it be just as likely 
to give a value greater  than 1 as belowand balance out on average?  Perhaps it is capped at one,
but then one would expect that the greater the amount of noise the farther off the result. However
int vs int and adv vs adv give the same result of .723544
 
Bob Koca

---

Recharge--play some free games. Win cool prizes too! Play It!