x
Chess - Play & Learn

Chess.com

FREE - In Google Play

FREE - in Win Phone Store

VIEW

Draw? Engine Disagrees

  • #1

    Why does the engine evaluate this position as -2 (advantageous for black)?

    I just learned the Lucena position and Philidor defense, and as far as I understand, this is a dead draw.

    Am I missing something?

     
    I let the engine run until depth 45 (!) and this is all it came up with:
     

     

  • #2

    It is a draw but engines don't think in human terms when evaluating. The extra pawn means the engine will assess black to have an advantage and the forced draw with best play is beyond its horizon. Unless they are equipped with endgame tablebases they will do this in drawn but unbalanced endings eg. Rook vs bishop with no pawns, or positions with impregnable fortresses.

  • #3
    Strangemover wrote:

    It is a draw but engines don't think in human terms when evaluating. The extra pawn means the engine will assess black to have an advantage and the forced draw with best play is beyond its horizon. Unless they are equipped with endgame tablebases they will do this in drawn but unbalanced endings eg. Rook vs bishop with no pawns, or positions with impregnable fortresses.

    Interesting!

    How then do we know if a position is drawn, or if it's just the engine being an engine, if we are using just...an engine? happy.png Is it possible?

  • #4

    There are tablebases which have been painstakingly put together (Nalimov, Lomonosov) where you can put up to 7 pieces on a board in any combination and they will tell you what the result will be with best play. Both are free to download I think. Some but not all engines are equipped with these and the ones that aren't give these mis-evaluations you describe. Of course a puny human like you or me can see a position like the one you posted and know straight away it's a draw because of things we have learnt. 

  • #5
    Strangemover wrote:

    There are tablebases which have been painstakingly put together (Nalimov, Lomonosov) where you can put up to 7 pieces on a board in any combination and they will tell you what the result will be with best play. Both are free to download I think. Some but not all engines are equipped with these and the ones that aren't give these mis-evaluations you describe. Of course a puny human like you or me can see a position like the one you posted and know straight away it's a draw because of things we have learnt. 

    Cool! Thank you.

  • #6

    There are general rules in R+P endgames, and exceptions to those general rules that help us to know even without an engine whether a position is drawn. 

    The simplest key to determining whether such a position is winning is whether the side with the extra pawn can control the queening square with his K when the pawn is on the 7th. I don't see any way to do that, and I don't see a way to get into one of the exceptional positions.

     

    Therefore, even without using an engine, I'd expect this game to be theoretically drawn. In other words, I'd expect this position to be drawn regardless of whose move it is. I don't see any way for Black to make progress.

  • #7

    Which engine?

    Current development versions of Stockfish, and all clones like ASMfish, remain stuck at 0.00 evaluation.

  • #8

    Tablebase... that is all I got to say and I agree with @pfren

  • #9
    pfren wrote:

    Which engine?

    Current development versions of Stockfish, and all clones like ASMfish, remain stuck at 0.00 evaluation.

    I tried Chess.com's Stockfish: https://www.chess.com/analysis-board-editor

    But even if you click on the analysis icon next to the first board I posted, you see the eval is -2.

     

    And thanks everyone for your input!

  • #10

    [Event "stadj vs. mormonchessman"] [Site " Chess.com"] [Date "Apr 10, 2018"] [White "stadj"] [Black "mormonchessman"] [Result "1/2-1/2"] [WhiteElo "1397"] [BlackElo "1451"] [TimeControl "1 in 0 day"] [SetUp "1"] [Termination "Game drawn by agreement"] 1. e4 e6 2. Nf3 Nf6 3. Nc3 d5 4. e5 Ng8 5. d4 c5 6. Bb5+ Nc6 7. O-O Qb6 8. Bxc6+ Qxc6 9. dxc5 Bxc5 10. Be3 Ne7 11. Bxc5 Qxc5 12. Qd4 Qxd4 13. Nxd4 a6 14. f4 O-O 15. a3 Bd7 16. b4 Rac8 17. Nd1 Rc4 18. c3 Ba4 19. Ne2 Nf5 20. Rf3 Rfc8 21. Nb2 Bb3 22. Nxc4 Bxc4 23. Nd4 Nxd4 24. cxd4 b5 25. Rc3 g6 26. Re1 Kg7 27. Ree3 h5 28. Kf2 f6 29. Rg3 f5 30. h4 Ra8 31. Kg1 a5 32. Kh2 axb4 33. axb4 Ra4 34. Ra3 Rxa3 35. Rxa3 Kf7 36. Ra7+ Kf8 37. g3 Be2 38. Kg1 Bd3 39. Kf2 Bc4 40. Ke3 Bf1 41. Kd2 Bc4 42. Kc3 Bf1 43. Rh7 Be2 44. Ra7 Bf1 45. Rc7 Be2 46. Rc6 Kf7 47. Rd6 Bf1 48. Ra6 Be2 49. Ra7+ Kf8 50. Kb3 Bc4+ 51. Ka3 Be2 52. Rc7 Bd3 53. Kb2 Be2 54. Kc1 Bf1 1/2-1/2 interesting draw

  • #11

    it is because for engines , there is no perpectual check rule when evaluating. as long as black has an extra pawn , he is ahead.

  • #12

    Stockfish  on my laptop evaluates this position at 0.00 withing one second and never changes. I do not have a tablebase 

  • #13
    DaniSpringer wrote:
    pfren wrote:

    Which engine?

    Current development versions of Stockfish, and all clones like ASMfish, remain stuck at 0.00 evaluation.

    I tried Chess.com's Stockfish: https://www.chess.com/analysis-board-editor

    But even if you click on the analysis icon next to the first board I posted, you see the eval is -2.

     

    And thanks everyone for your input!

    That chess.com analysis is very shallow depth 20 ply(useless,). Install SF in your computer you must have 30+ ply to get proper analysis .. In some positions, SF should reach 50 ply+ for analysis.

  • #14
    mickynj wrote:

    Stockfish  on my laptop evaluates this position at 0.00 withing one second and never changes. I do not have a tablebase 

    Yeah. Pfren also pointed that out. @Mike_Arunchuk's comment about the engine sounds antiquated to me. Sure, there are positions that engines misevaluate. But this isn't one of them.

  • #15

    A properly installed Stockfish (with table bases) can even announce Mate in 80 in some positions .

    https://www.chess.com/forum/view/general/how-strong-is-your-stockfish-i-bet-90-of-people-dont-have-6-men-tb

  • #16
    DaniSpringer wrote:
    Strangemover wrote:

    It is a draw but engines don't think in human terms when evaluating. The extra pawn means the engine will assess black to have an advantage and the forced draw with best play is beyond its horizon. Unless they are equipped with endgame tablebases they will do this in drawn but unbalanced endings eg. Rook vs bishop with no pawns, or positions with impregnable fortresses.

    Interesting!

    How then do we know if a position is drawn, or if it's just the engine being an engine, if we are using just...an engine?  Is it possible?

    Check with a tablebase.

    Or ask a human. (Preferably an Expert of Master.)

    Computers can do many amazing things in chess. But endgames still belong to us biologic life-forms. 

  • #17

    I don't know what was wrong with the engine the OP was using, but Komodo or Stockfish 8, running on my three-year old laptop, evaluate the position at 0.00 in less than 5 seconds

  • #18
    SmyslovFan wrote:

    There are general rules in R+P endgames, and exceptions to those general rules that help us to know even without an engine whether a position is drawn. 

    The simplest key to determining whether such a position is winning is whether the side with the extra pawn can control the queening square with his K when the pawn is on the 7th. I don't see any way to do that, and I don't see a way to get into one of the exceptional positions.

     

    Can you elaborate on that please? After, for example 1 Rf8 Re7 2 Ra8 Rg7+ 3 Kh2 f2 Black's pawn is on the 7th and his King controls the queening square. It's still a draw because White now checks from the side, but this seems to break your rule.

  • #19
    IpswichMatt wrote:
    SmyslovFan wrote:

    There are general rules in R+P endgames, and exceptions to those general rules that help us to know even without an engine whether a position is drawn. 

    The simplest key to determining whether such a position is winning is whether the side with the extra pawn can control the queening square with his K when the pawn is on the 7th. I don't see any way to do that, and I don't see a way to get into one of the exceptional positions.

     

    Can you elaborate on that please? After, for example 1 Rf8 Re7 2 Ra8 Rg7+ 3 Kh2 f2 Black's pawn is on the 7th and his King controls the queening square. It's still a draw because White now checks from the side, but this seems to break your rule.

    Let me show you with a diagram:

     

  • #20

    Which engine gives wrong evaluation ? Just checked it with asmfish and shows 0.00 from very start.

Online Now