As good as Quackle is in general, it has a known bug in the endgame.  I recall 
reading an email on the issue, wherein the valuation of moves in the final 
stages was totally wrong.  So it would appear that unfortunately this current 
discussion is purely academic...
Trevor

To: [email protected]
From: [email protected]
Date: Sat, 8 May 2010 16:33:44 -0700
Subject: Re: [quackle] Re: seeing possible future moves


















 



  


    
      
      
      


On Sat, May 8, 2010 at 3:14 PM,  <[email protected]> wrote:
















 



  


    
      
      
      Dear Matt,



With one tile in the bag, there are only eight racks opponent could have, 

so each candidate play should have a win percentage of some multiple of 

12.5%. The only exception should be ties (or the extremely rare candidate play 
of 

passing).



All the best,



Jim Kramer</HTML>  


Forgive my kibitzing - I lurk here out of theoretical interest only at this 
point in time.

What Jim and Albert say would certainly be true for an exhaustive-search 
end-game player.  A simulation that only looks 1 or 2 moves deep would not be 
expected to give such clean results for an endgame with a deeper move tree than 
that.


When David states he "simulated" in the the original question, I assume that 
simulation does not do exhaustive search in the endgame (even if the 
"Championship Player" might).  Forgive me if that assumption is wrong, I only 
state it because it might be the source of the disagreement.


Steven Gordon  





    
     

    
    




    
     

    
    






                                          
_________________________________________________________________
Browse profiles for FREE! Meet local singles online.
http://clk.atdmt.com/NMN/go/150855801/direct/01/

Reply via email to