Yeah, yeah, duplicate... Dupe A Dupe B ...but the times they are a-changing. For example, I tried this "rambling rook" piece by Henneberger, Lichess Stockfish found a #31 on depth 93 or so. The famous Nxb6 trap (can't find it offhand) has long been avoided with deep enough depth (White instead gangs upon Pa5 and wins slowly). The "Penrose" now also gets an immediate yawning "0.0" by Stockfish.

The question probably breaks down to which depth the super modern engines can reach, and if a problem can have its punchline beyond this horizon. (E.g., any "Penrose" style fortress problem will only need a depth of 100 halfmoves, when the 50 move rule kicks in, so this won't impress a modern engine.)

  • 4
    Yes. The starting position...
    – SmallChess
    Sep 2 '21 at 12:50
  • If you already know its a "Duplicate" with Dupe A Dupe B , why post? Short Answer is No... Alpha Zero works differently to other engines, so after an initial "viewing" of n amount of games. It will use its AI just to kick something off. Sep 2 '21 at 14:02
  • @DavidBateman how does AlphaZero working differently detract from this question? AlphaZero style engines aren't even good at all tactics, for instance in a correspondence game I had a position where Stockfish would find a tactical solution but Leela would not. (but since Stockfish did find it, that doesn't really help here)
    – koedem
    Sep 2 '21 at 21:28
  • In case you are interested though, I have not tried recently so maybe Leela can find it now, but not the last time I tried. White to play and win: r5r1/1np3bk/1p1p4/3P4/bpP2PPp/P3N2P/1q1BQ2K/4R1R1 w - - 0 33
    – koedem
    Sep 2 '21 at 21:33
  • @SmallChess unsolvable for computers but solvable for humans?
    – BCLC
    Sep 2 '21 at 22:57

Browse other questions tagged or ask your own question.