Position upper bound 169 bits WITH PROOF

588 views
Skip to first unread message

Full Decent

unread,
Sep 5, 2006, 1:26:07 PM9/5/06
to
Hello, I have not found any proofs for the upper bound on the number of
chess positions, so here is mine, clocking in at 169 bits.

Will Entriken - 2006
Encoding chess positions with about 164 bits.


This paper describes how to encode any legal chess position in less
than 169 bits. It appears that this approach has not been considered,
so I am publishing it.


==Denote the armies==

Each player has an army that may be written as :

<- A -> <- B ->
K P P P P X X X X Q R R N N B B

Where
- X may be one of [QRNB]
- A + B <= 8
- Any peice starting with X is optional

We will be looping through every possible combination of each players'
armies. While doing this, we will keep a note of the following for each
army:
- A = The number of pawns (since pawns can only be placed on the 8x6
rectangle)
- S = The size of the army (to determine the number of empty squares)
- P = The product of the factorial of the number of each type of
piece (how indistinguishable is the set?)


==Placement==

For any two armies, we have [A, S, P, a, s, p] and are able to
calculate how many ways those pieces can be placed on the board. This
number is: 48! * (64-A-a)! / (48-A-a)! / (64-S-s)! / P / p.

48! (64-A-a)!
--------- * ---------
(48-A-a)! (64-S-s)!

---------------------

P * p


==Results==

Possible positions: 23937533792747905898433845980097921846050276105440
log_2: 164.033
log_10: 49.379

Therefore, given a chess position, this method can assign a number
between 1 and 239...440 to it. This is reversable as well.


==Extras==

I did not count positions multiple times when it is ambiguous whether
any castling is possible. Also, certain positions may have up to 5 en
passantable pawns. These ineficiencies cost an extra:

Possible positions: 24 (factor)
log_2: 4.584
log_10: 1.380

White is always to act. If black is to act, invert the board. Those
positions are the same.


==Conclusion==

Therefore we now have a fast, reversible way to encode a position into
169 bits and a mathematically proven upper bound on the number of chess
positions.


==Appendix==

Thjis is the counting program I used. It requires the GNU GMP library
and compiles with gcc -o ub -lgmp ub.c

#include <stdlib.h>
#include <stdio.h>
#include <gmp.h>

int fact (int i)
{
return (i<=1) ? 1 : i*fact(i-1);
}

int main()
{
int base_army[5] = {0,1,2,2,2}; /* Armies attainable without
pawns or promotion */
int army[5]={0}; /* PAWNS, QUEENS, ROOKS,
BISHOPS, KNIGHTS */
int ab; /* Number of pawns + promoted
pieces */
int s; /* Size of the army */
int p; /* Prod(Fact(Count(each
peice))) */

int ARMY[5]={0}; /* ... and for the other player
*/
int AB;
int S;
int P;

int i;
mpz_t facts[65]; /* precompute fact(0..64) */
mpz_t total;
mpz_t current;

mpz_init(total);
mpz_init(current);

for(i=0; i<=64; i++)
{
mpz_init(facts[i]);
mpz_fac_ui(facts[i], i);
}

for(army[0]=0; army[0]<=8; army[0]++)
for(army[1]=0; army[1]<=9; army[1]++)
for(army[2]=0; army[2]<=10; army[2]++)
{
for(army[3]=0; army[3]<=10; army[3]++)
for(army[4]=0; army[4]<=10; army[4]++)
{
ab = 0;
s = 1; /* Count the King */
p = 1;

if (army[0]+army[1]+army[2]+army[3]+army[4] > 16)
continue; /* Quick impossible check */

for (i=0; i<5; i++)
{
if (army[i] > base_army[i]) ab += army[i] -
base_army[i];
}
if (ab > 8)
continue; /* Impossible army */

for (i=0; i<5; i++)
{
s += army[i];
p *= fact(army[i]);
}

for(ARMY[0]=0; ARMY[0]<=8; ARMY[0]++)
for(ARMY[1]=0; ARMY[1]<=9; ARMY[1]++)
for(ARMY[2]=0; ARMY[2]<=10; ARMY[2]++)
for(ARMY[3]=0; ARMY[3]<=10; ARMY[3]++)
for(ARMY[4]=0; ARMY[4]<=10; ARMY[4]++)
{
AB = 0;
S = 1;
P = 1;

if (ARMY[0]+ARMY[1]+ARMY[2]+ARMY[3]+ARMY[4] >
16)
continue;

for (i=0; i<5; i++)
{
if (ARMY[i] > base_army[i]) AB +=
ARMY[i] - base_army[i];
}
if (AB > 8)
continue; /* Impossible army */

for (i=0; i<5; i++)
{
S += ARMY[i];
P *= fact(ARMY[i]);
}

/*
printf("army[]=%d%d%d%d%d -- ",
army[0],army[1],army[2],army[3],army[4]);
printf("ARMY[]=%d%d%d%d%d -- ",
ARMY[0],ARMY[1],ARMY[2],ARMY[3],ARMY[4]);
printf("a=%d, s=%d, p=%d\tA=%d, S=%d, P=%d\n",
army[0], s, p, ARMY[0], S, P);
*/


mpz_set_ui(current, 1);
mpz_mul(current, current, facts[48]);
mpz_divexact(current, current, facts[48 -
army[0] - ARMY[0]]);
mpz_mul(current, current, facts[64 - army[0] -
ARMY[0]]);
mpz_divexact(current, current, facts[64 - S -
s]);
mpz_divexact_ui(current, current, P);
mpz_divexact_ui(current, current, p);
mpz_add(total, total, current);
}
}
printf("%d/990\n", army[0]*10*11 + army[1]*11 + army[2]);
mpz_out_str(stdout, 10, total);
puts("");
}
mpz_out_str(stdout, 10, total);
puts("");
return 0;
}

Full Decent

unread,
Sep 5, 2006, 1:45:32 PM9/5/06
to
A quick correction, 24 above should be 6*2^4 = 96, adding 2 bits

Anders Thulin

unread,
Sep 5, 2006, 2:35:04 PM9/5/06
to
Full Decent wrote:
> Hello, I have not found any proofs for the upper bound on the number of
> chess positions,

Published in some old issue of ICCA, I think. 1991?

Those that have been mentioned in the groups seem to end up around 160 bits.
(See old postings), but there have been quite a number of ideas how to
decrease that in various circumstances.

--
Anders Thulin ath*algonet.se http://www.algonet.se/~ath

Full Decent

unread,
Sep 16, 2006, 12:05:36 PM9/16/06
to

I found out that the article you are referring to is in Sept 1996 of
ICCA, but I can't find that journal anywhere.

Do any of these people claiming 160bits have proof?

v3ng...@gmail.com

unread,
May 17, 2014, 12:38:23 PM5/17/14
to
Actually - I feel if you broke the board into two sets, one of 48 squares and one of 64, you could then use base-n to describe the pieces...
So for the pawns you have 3 states, which are an empty square, a white pawn or a black pawn, so that is 48 to the power of 3, or 110,592 combinations...
For the other 5 types of pieces there are also 2 colors, and empty squares, they can exist on 64 squares, so again it is 64 to the power of 11 (5*2 + the empty square) - this yields 73,786,976,294,838,206,464 states, multiplying the two you get a number which can be stored in 83-bits... my math could be out - but I'd be more inclined to follow this approach...
(this would only be the physical board state, not the castling status, or any other information...)

Richard Delorme

unread,
May 21, 2014, 7:41:01 AM5/21/14
to
Le 17/05/2014 18:38, v3ng...@gmail.com a �crit :
> Actually - I feel if you broke the board into two sets, one of 48
> squares and one of 64, you could then use base-n to describe the
> pieces... So for the pawns you have 3 states, which are an empty
> square, a white pawn or a black pawn, so that is 48 to the power of
> 3, or 110,592 combinations...

No, this is 3 to the power of 48. But this include too many cases, with
up to 48 black pawns or 48 white pawns.

> For the other 5 types of pieces there are also 2 colors, and empty
> squares, they can exist on 64 squares, so again it is 64 to the power
> of 11 (5*2 + the empty square) - this yields 73,786,976,294,838,206,464
> states, multiplying the two you get a number which can be stored in
> 83-bits... my math could be out - but I'd be more inclined to follow
> this approach... (this would only be the physical board state, not the
> castling status, or any other information...)

Same here, this is 11 to the power of 64.


v3ng...@gmail.com

unread,
Jun 1, 2014, 10:03:50 AM6/1/14
to
Aye I realised my error when I tried to implement, good call, and outstanding post by the O.P.

fulld...@gmail.com

unread,
Jul 7, 2014, 7:22:41 PM7/7/14
to
FYI, there is some light discussion on the topic at https://codegolf.stackexchange.com/questions/19397/smallest-chess-board-compression that tries approaches used Huffman encoding.
Reply all
Reply to author
Forward
0 new messages