[rc5] v3

Sebastian Kuzminsky kuzminsk at taussky.cs.colorado.edu
Tue Oct 28 06:42:29 EST 1997


Joseph Fisk <joe at rm405n-a.roberts.fit.edu> wrote:
] >    But if each board takes 200 bits to store, that's tens of thousands
] > of terabytes.  How do they do it?  Is there a computer chess programmer
] > in the house?
] 
] You could store each possiblity (every single combination wouldn't have to
] be stored; those that lead immediately to checkmate, for example =) as
] only a few bits describing the change from the previous state.  Also
] (6*10^14)*200bits = 15000000000000000 terabytes. =)


   Hm, 6 * 10^14 * 200 is 1.2 * 10^17 bits, which is 1.5 * 10^16 bytes,
which is about 1.5 * 10^13 KB, or 1.5 * 10^10 MB, or 1.5 * 10^7 GB, or
1.5 * 10^4 TB:  tens of thousands of terabytes.




] There is bound to be a lot of repition there; zip would probably take it
] down significantly although obviously that many terabytes is pretty
] extreme.  hehehe..


   Agreed, there is a lot of redundancy in the 10-ply search tree
described above.


   Note that while it takes about 200 bits to describe a board, it takes
no more than 9 bits to describe a move, and often less:  4 or less bits
to name the piece, and then 5 or less bits to name the destination
square.  Transmitting large search trees trees thus takes only 1/20 of
the bandwidth if you submit moves (think of them as diffs!) instead of
boards.




Sebastian

----
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.



More information about the rc5 mailing list