Can chess be analytically "solved" in the forseeable future?

Brian-E · 2012-04-02, 23:12

It is normally assumed that it is not feasible to "solve" chess, that is to determine beyond any doubt what the result of the game is if both players play perfectly from the starting position, and what the "best" move(s) is/are from any arbitrary position.

This can and has been done for endgame positions with up to about 7 pieces on the board, I believe. So-called table bases of all possible positions together with the result of the game with best play are generated using retrograde analysis, and the number of positions with this reduced number of pieces is small enough for that to be possible. But with 32 bits on the board at the starting position and some 20 legal moves to start with, rising to 40 or more legal moves for each player in more general positions, the number of possible continuations becomes astronomical, as is well known.

But could it be feasible to analyse chess out after all? This article gives an interview with a someone who claims to have analysed a certain opening, the Kings Gambit Accepted, 1.e4 e5 2.f4 exf4 to a draw if White plays 3.Be2, or otherwise a win for Black. Is the Kings Gambit Accepterd position, after just two moves by each player, many orders of magnitude less complicated than the starting position? If not, does this mean that it may become feasible after all to fully analyse the game of chess so that "best" moves and the result with "best" play from the starting position and any other position can be known?

I would be very interested to read any views about this.

Zeta-Flux · 2012-04-02, 23:53

Brian-E,

This is one of my interests. The current state of affairs is that all 6 piece endgames are solved (modulo some silly positions, like 5 pieces versus 1 king, or positions where castling is still available). The storage space for all these positions is more than 1 terabyte (which nowadays isn't too bad!).

There are certain 7 piece endgames that have been solved, but this is not complete (or at least not available publicly). There is one 7 man endgame that takes 517 moves (ignoring the 50 move rule) to be forced to convert to fewer pieces.

This brings up two problems in such searches: do you ignore the 50 move rule? what metric do you use to measure a position (moves to conversion? moves to mate? win/loss? etc...)? Depending on the metric you use, you get different moves as the best possible. The 6 men tablebases use all metrics (if I remember right).

What was really interesting about the article you linked to is that they don't try to definitely prove mate, they simply work out lines until the advantage is overwhelming. Apparently, the king's gambit accepted is so "sharp" that it allows computers to compute lines always to overwhelming advantages. The problem with general tablebases is that the positions often are not overwhelming.

So, to answer you question about whether chess will be solved: the answer could be yes if there is a sufficiently sharp starting move, that allows perfect play to remain overwhelming. Of course, this won't "prove" that chess is solved, it will only be a probabilistic argument. And it won't "strongly" solve it (which would tell you how to respond to every possible opening).

But that article is very cool. Thank you for linking it.

Zeta-Flux · 2012-04-03, 00:00

By the way, I've always wanted to "solve" the position 1. e4 e5 2. Nf3 f6. I wonder if this program could do that.

KyleAskine · 2012-04-03, 00:36

I assume this is an April Fools Joke. Chessbase always has one. Even though it says April 2nd, they claim in the article the day they talked was the 1st.

cheesehead · 2012-04-03, 01:20

Quote:

Originally Posted by KyleAskine

I assume this is an April Fools Joke. Chessbase always has one.

But Chessbase (http://www.chessbase.com/index.asp) has another article, dated exactly 01.04.2012, which looks far, far more suspicious than the King's Gambit article:

"The 'Let's Check' Crystal Ball"
www.chessbase.com/newsdetail.asp?newsid=8045

Quote:

Even though it says April 2nd, they claim in the article the day they talked was the 1st.

They claim that the day on which the interview was agreed to was April 1. It could have taken place on April 2. :-)

Zeta-Flux · 2012-04-03, 03:27

I have to say, I hate April Fools. They got me on this one.

KyleAskine · 2012-04-03, 11:11

Quote:

Originally Posted by cheesehead

But Chessbase (http://www.chessbase.com/index.asp) has another article, dated exactly 01.04.2012, which looks far, far more suspicious than the King's Gambit article:

"The 'Let's Check' Crystal Ball"
www.chessbase.com/newsdetail.asp?newsid=8045

I don't get it. I own Fritz 13. Let's Check is a real thing. I feel like I am missing a joke.

By the way, it's great that we seem to have so many chess fans on this board! I am Executive Director of the DC Chess League, so I have always been heavily involved in the chess scene for a while.

Xyzzy · 2012-04-03, 18:23

Quote:

By the way, it's great that we seem to have so many chess fans on this board!

Something to consider, if you have the time, is introducing Boy Scouts to the wonders of chess by volunteering to be a chess merit badge counselor.

http://www.scouting.org/sitecore/con.../mb-CHESS.aspx

The curriculum was developed with input from the USCF.

http://main.uschess.org/content/view/11348/639

Cub Scouts also have a chess belt loop and pin that they can earn.

http://www.scouting.org/sitecore/con...nda/chess.aspx

Brian-E · 2012-04-04, 08:15

Quote:

Originally Posted by Zeta-Flux

I have to say, I hate April Fools. They got me on this one.

Yes, I hate them too. They got me too. In this article they have come clean about it now. Makes me feel a bit silly.

Anyway, it's generated some interesting posts here. And if someone like you, Zeta-Flux, did not reject the possibility out of hand that a playable opening like the King's Gambit could have been analysed completely to a (probabilistically) determinable result, then maybe it is still possible in the forseeable future.

I appreciate what you say about the necessity of the opening being "sharp" if it is to be fully analysed. That's mentioned in the spoof article too. I guess the starting position, which is essentially "quiet" because no pieces have been developed, is not amenable to analysis in that way. Nor is a quiet flank opening any good for doing that.

I used to play correspondence chess. I've given up now because it's been spoiled by the general use of chess playing software to help players make their moves, and I had no desire to use such software myself to help me find the right moves. In the last few years that I used to play it, I always chose quiet positional openings and similarly positional continuations, always trying to reach an endgame quickly and avoiding tactical middlegames, precisely because quiet positions lend themselves less well to analysis by software.

cheesehead · 2012-04-05, 19:16

Quote:

Originally Posted by Zeta-Flux

I have to say, I hate April Fools. They got me on this one.

April Fools Day is my least favorite 'holiday". My former wife liked to pull practical jokes on me that day every year; I wasn't amused.

science_man_88 · 2012-09-26, 21:53

one thing popped up in my mind today:

can progressive chess be solved even if normal chess can't ? , the reason I ask is because if it can we have a backward way to solve normal chess I think. I think this because normal chess is basically progressive chess with only one non reversed move/capture per player's turn.

2012-04-02, 23:12	#1
Brian-E "Brian" Jul 2007 The Netherlands 7·467 Posts	Can chess be analytically "solved" in the forseeable future? It is normally assumed that it is not feasible to "solve" chess, that is to determine beyond any doubt what the result of the game is if both players play perfectly from the starting position, and what the "best" move(s) is/are from any arbitrary position. This can and has been done for endgame positions with up to about 7 pieces on the board, I believe. So-called table bases of all possible positions together with the result of the game with best play are generated using retrograde analysis, and the number of positions with this reduced number of pieces is small enough for that to be possible. But with 32 bits on the board at the starting position and some 20 legal moves to start with, rising to 40 or more legal moves for each player in more general positions, the number of possible continuations becomes astronomical, as is well known. But could it be feasible to analyse chess out after all? This article gives an interview with a someone who claims to have analysed a certain opening, the Kings Gambit Accepted, 1.e4 e5 2.f4 exf4 to a draw if White plays 3.Be2, or otherwise a win for Black. Is the Kings Gambit Accepterd position, after just two moves by each player, many orders of magnitude less complicated than the starting position? If not, does this mean that it may become feasible after all to fully analyse the game of chess so that "best" moves and the result with "best" play from the starting position and any other position can be known? I would be very interested to read any views about this.

2012-04-02, 23:53	#2
Zeta-Flux May 2003 3013₈ Posts	Brian-E, This is one of my interests. The current state of affairs is that all 6 piece endgames are solved (modulo some silly positions, like 5 pieces versus 1 king, or positions where castling is still available). The storage space for all these positions is more than 1 terabyte (which nowadays isn't too bad!). There are certain 7 piece endgames that have been solved, but this is not complete (or at least not available publicly). There is one 7 man endgame that takes 517 moves (ignoring the 50 move rule) to be forced to convert to fewer pieces. This brings up two problems in such searches: do you ignore the 50 move rule? what metric do you use to measure a position (moves to conversion? moves to mate? win/loss? etc...)? Depending on the metric you use, you get different moves as the best possible. The 6 men tablebases use all metrics (if I remember right). What was really interesting about the article you linked to is that they don't try to definitely prove mate, they simply work out lines until the advantage is overwhelming. Apparently, the king's gambit accepted is so "sharp" that it allows computers to compute lines always to overwhelming advantages. The problem with general tablebases is that the positions often are not overwhelming. So, to answer you question about whether chess will be solved: the answer could be yes if there is a sufficiently sharp starting move, that allows perfect play to remain overwhelming. Of course, this won't "prove" that chess is solved, it will only be a probabilistic argument. And it won't "strongly" solve it (which would tell you how to respond to every possible opening). But that article is very cool. Thank you for linking it. Last fiddled with by Zeta-Flux on 2012-04-02 at 23:53

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Stockfish game: "Move 8 poll", not "move 3.14159 discussion"	MooMoo2	Other Chess Games	5	2016-10-22 01:55
Aouessare-El Haddouchi-Essaaidi "test": "if Mp has no factor, it is prime!"	wildrabbitt	Miscellaneous Math	11	2015-03-06 08:17
"Cannot initialize FFT code" with no swap (SOLVED)	Unregistered	Information & Answers	0	2010-07-15 14:10
Would Minimizing "iterations between results file" may reveal "is not prime" earlier?	nitai1999	Software	7	2004-08-26 18:12

2012-04-03, 00:00	#3
Zeta-Flux May 2003 7×13×17 Posts	By the way, I've always wanted to "solve" the position 1. e4 e5 2. Nf3 f6. I wonder if this program could do that.

2012-04-03, 00:36	#4
KyleAskine Oct 2011 Maryland 2×5×29 Posts	I assume this is an April Fools Joke. Chessbase always has one. Even though it says April 2nd, they claim in the article the day they talked was the 1st.

2012-04-03, 03:27	#6
Zeta-Flux May 2003 7×13×17 Posts	I have to say, I hate April Fools. They got me on this one.

2012-09-26, 21:53	#11
science_man_88 "Forget I exist" Jul 2009 Dumbassville 8384₁₀ Posts	one thing popped up in my mind today: can progressive chess be solved even if normal chess can't ? , the reason I ask is because if it can we have a backward way to solve normal chess I think. I think this because normal chess is basically progressive chess with only one non reversed move/capture per player's turn.