Chess hasn't been completely solved yet, even though it is completely deterministic. However, can't we "solve" it up to a certain number of moves (e.g., up to 50 moves)? Then, considering perfect play by both players, who would win, or would it be a draw?
1
-
reddit.com/r/chess/comments/1cy94fi/…– ARGYROU MINASCommented May 25 at 11:07
Add a comment
|
1 Answer
It's hard to start from the opening position, as then you don't know what the result of the line is.
Instead, chess is "being solved" from the end, with endgame tablebases.
All positions of 7 pieces (including the kings) have been completely solved, there are in databases in which you can look up the result in case of perfect play. There has been some work on 8 pieces, but the required investment in computer power is too large -- it's expected to take about 2 petabyte of space when finished. ( Wikipedia on tablebases ).
Let's say we would start from the front. How many different ways to play n moves are there? The perft page on the chess programming Wiki has a list.
Let's say we store 1 byte per line, and the 2 petabyte of the 8 piece tablebase is a limit, then we can store about 2 million billion lines. That's roughly the perft for 11 half-moves.
So in the same space as 8 piece tablebases, very roughly, you could get up to all lines after the 6th move of white.
But the problem of starting from the beginning remains -- you can't conclude anything about those positions being a win, loss or draw.
-
"you can't conclude anything about those positions being a win, loss or draw" unless you can look at the results in a very large database of games and use machine learning. But that needs the word "calculated" to be used in a different way to the what I expect the question asker had in mind. Commented May 25 at 13:45
-
That would especially need the word "solved" to be used differently, as that generally implies mathematical proof. You can get an engine evaluation of course, but that's not proof. Commented May 25 at 20:35