lichess.org

If we solve chess, white can force win

White not in zugzwang so it's unclear
0.5 move advantage is not big but white have edge at starting position

What if Black play's for a draw?

Can black make a fortress, that causes a draw on every game ?
Is so, white will not be able to force a win, even if chess is solved.

Set an engine to draw more than win and test out the theory.
Discourage White from accepting a draw by increasing it's positive contempt centipawn number.
With the Black pieces use another engine and decrease it's contempt to increase draw possibilities.

Then you might see that white will have a hard time winning. White will be fighting an opponent that is happy with draws.

This method might even show the main line to use ... to always want to win games when playing the white pieces.

what exactly does solving chess mean? I assumed it meant whatever moves your opponent makes the outcome will always be win, or the outcome will always be a draw, depending on the outcome of solving chess.

ie if chess is solved to be a win for white then black making a fortress won't make a difference.

But I've never looked up the definition of what solving chess means.

Weak/soft solution: has been resolved for say 20 years, humans can go home

Hard/complete solution: a 32-men tablebase. In the next couple of decades probably 10 pieces if nothing extraordinary happens. Nothing more.

Chess is not weakly solved.

Whether it's a win, a draw or a loss for white is not even proven (though it's obvious it's a draw).

BTW you can find 7 pieces solved endgame tablebase here:

syzygy-tables.info/ (17 TB)

Syzygy tablebases allow perfect play with up to 7 pieces, both with and without the fifty-move drawing rule, i.e., they allow winning all won positions and bringing all drawn positions over the fifty-move line.

Here is the Lomonosov. (140 TB)

http://tb7.chessok.com/

Those took about 6 months to solve using the Lomonosov supercomputer.

the 8 piece tablebase would be 1000 TB in size.


the record number of digits of pi that someone remembers is now 67 000. of course chess has several lines, not just one. but it shows the vast capability of the human brain. or some brains...

the invention of cars did not make it less interesting to have or watch physical exercise

I stand corrected. Thought of something like "de facto" solved in respect to humans.

the games that game theory considers can all be included in a the bigger set of all legal games (which i call 42_)

that means all games: the blundering ones (which game theory does not consider, i believe, tell me wrong), and the zero random random games). the game theory idealized player, rational and of infinite memory and horizon, its moves are also included.

I will make it even bigger by developing the draw endpoints into pat, and n-repeats (no max move termination).

This is intended to put the draw apparent imbalance that this thread has referred to, in terms of probable forced games, into a fairer perspective in terms of a priori end-points (rules of termination should not be imbalanced from the start).

n-repeat behaviors for all possible n (which would become an interesting question, what is the maximum n). The board being finite and the number of pieces as well, i ruled out ergodic sequences (that would be infinite pattern repeat).

This would make for a fairer comparison of termination classes: WW, BW, WP, BP, W n-repeat (n=1,...., maxn).

So the 50 max draws should be excluded when making statistics on the best games (closest to game theory assumption), or start making bins for n-repeats.

How many pats, compared to win, and lose?

42_=W+D+L
W=WW+BW
L=WL+BL (just for flexibility of syntax, since L=W, WL=BW, WW=BL)
D=WP+BP
nRep= WnRep+BnRep, n from 1 to whatever.

are each of the separated classes that would stem from this particular end-point classification be more balanced in term of their frequency in the set of all possible legal games*(-50rule). competitie, cooperative, and plain random included?


You can't post in the forums yet. Play some games!

Reconnecting