The way the puzzle system is designed, there can only be one solution and many endgames with less than 8 pieces, where tablebase are available, will have multiple lines. That's especially true since the site uses Syzygy tablebases which don't show the fastest wins.
Chess.com should generate puzzles using endgame databases
Sort:
Martin_Stahl wrote:
The way the puzzle system is designed, there can only be one solution and many endgames with less than 8 pieces, where tablebase are available, will have multiple lines. That's especially true since the site uses Syzygy tablebases which don't show the fastest wins.
But my understanding is they scrape through every calculated game and only select for puzzles the set of moves which have only one "good" line.
Can't they do the same thing with endgame databases?
So as we know, chess.com primarily generates it's puzzles via moves in chess.com games.
The reason it does this rather than just using random piece generators is, I assume:
-So the piece arrangements are "realistic" to real games.
-Because the games are already calcated and they can use the calculated value of moves to decide how the puzzles go.
But I've been thinking, there is a very high proportion of possible wndgames which are "realsitic" to achieve in actual games. And endgame databases already "calculate" every possible endgame. So why not use those?
You'd only need a few simple rules to ensure the games are realistic. The ones I'd suggest are:
-There is at least one pawn on the board
-The position must be obtainable without pawn promotion.
-There are no tripled, quadrupled etc pawns.
-From the perspective of white, white pawns are always behind any black pawns in the same file.
-From the perspective of white, the white king is no further advanced than the black king.
-No piece is in a corner, except rooks.
_______________
What do we think of this idea?