Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Compact encoding of chess positions

Author: Reinhard Scharnagl

Date: 07:07:53 03/05/05

Go up one level in this thread


On March 05, 2005 at 08:10:27, Chris Welty wrote:

>On March 05, 2005 at 02:11:33, Reinhard Scharnagl wrote:
>
>>On March 04, 2005 at 14:45:46, Chris Welty wrote:
>
>>>I tried it with a slightly different restriction: Each side may have no more
>>>than 7 officers (officer=Q/R/B/N). An upper bound on the number of positions
>>>with this restriction is 2.3754e+043 and it can therefore be encoded in 144.091
>>>bits. Probably this can be reduced by another 1-2 bits by someone really
>>>determined.
>>>
>>>Is a 7-officer maximum realistic in actual games?
>>
>>Hi Chris,
>>
>>a) if one player is superior to the other in practical games there could be
>>easyly raised more the seven officers in the piece set of the better player.
>>
>>So I think that such a restriction would be somehow unrealistic.
>
>In real games I would have thought checkmate would occur before 8 officers - but
>I'd love to see a real game where one side had more than 7.

Well, you start with seven officers. Promoting once and mating then will do.

Think of a constellation with one very bad and one very good player.

>>b) Of course you can conclude from a maximum count of positions to an existing
>>encoding length when providing such a huge look up table for all existing chess
>>positions. But what I experimented with has been a realistic encoding scheme
>>without such an utopic look up table.

>A "utopic look up table" is quite realistic. That is how Nalimov Tablebases
>work.

I have been talking on a table with 2*(10^43)*172 Bits what means to encode
about 4*(10^45) Bits. I do not know where you will find such a large memory.
Think of a magnitude like to store a Bit in every atom of the moon.

Reinhard.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.