How strong can the smartest chess engine in the world play intuitive chess?

Posts : 880 Join date : 2020-11-25 Location : USA

Intuitive chess is where the chess engine decides on a move with no calculation. But plays based on positional understanding only.
Lc0 is by far the smartest chess engine I have tested. With this kind of play.

Lc0 vs Stockfish at intuitive chess. Lc0 wins 100% of the games.

I define intuitive chess as setting the chess engines to play a move at only 1 node.

When playing a match of 50 games. Lc0 was able to score 2 draws with just intuitive chess play. Against Stockfish DEV playing with at time control of 1m+1s.

And Lc0 was able to score a accuracy of .45 centipawns in the 50 games playing against Stockfish DEV. A accuracy just below GM level of play at 3 minutes a move at .40. And Lc0 is doing so with zero tactical calculation or look ahead.

Lc0 v0.29.0: 53/55/24/72/48/65/56/37/48/55/81/83/28/36/23/55/50/80/54/41/45/33/53/82/49/58/40/49/42/34/30/4/69/28/38/26/20/23/40/27/69/18/4/37/27/38/32/1/69/45 => Average=0.45

Here is the accuracy score of GM Magnus Carlsen last 50 bullets games played at 1m+1s on Chess.com. And GM Carlsen is achieving this accuracy with 2 mins of thinking time in a 60 move bullet game.

MagnusCarlsen: 60/54/68/51/44/71/38/29/55/49/28/27/68/31/51/77/25/23/69/18/54/57/20/28/49/11/29/47/58/16/33/16/59/26/8/25/14/10/11/18/8/6/39/20/42/14/24/37/13/14 => Average=0.38

Here are some other scores that were GM Magnus Carlsen's opponents at 1m+1s time controls.

Hikaru: 68/63/12/55/16/8/20/42/37 => Average=0.38
Duhless: 55/28/49/51/12/76/25/55/37/63/11/65/12/23/14/75/61 => Average=0.41
LyonBeast: 56/34/23/8/86/35/35/84 => Average=0.43
FabianoCaruana: 74/46/31/30/46/93/71/8/25/47 => Average=0.50
NikoTheodorou: 70/46/71/42/71/114 => Average=0.66

It is amazing that chess can be played at this level by just looking at the chess position. And it is very depressing for us human chess players. Who thought we knew more about positional understanding then the chess engine. And that is true, except for Lc0.

Here is Lc0's 2 draws playing Stockfish with intuitive play vs. Stockfish DEV playing with 1m+1s on 16 CPU's.

[pgn][Event "banksia game"]
[Site "*"]
[Date "2023.01.06"]
[Round "1"]
[Board "43"]
[White "Lc0 v0.29.0"]
[Black "Stockfish dev-20230104"]
[Result "1/2-1/2"]
[ECO "A35"]
[PlyCount "92"]
[EventDate "2023.??.??"]

1. c4 Nf6 2. Nf3 c5 3. Nc3 Nc6 4. g3 {A35: English, symmetrical, four knights system} e6 {Stockfish dev-20230104-9fe9ff00:} 5. Bg2 Qb6 6. O-O Be7 {End of opening} 7. e4 {+0.2/1 2212 1 284/567/149} d6 {+0.3/31 3990 57214367 1/963/36. OK} 8. Rb1 {10 +0.2/1 1 299/535/166. OK} O-O {+0.3/31 6446 92176550 1/965/34} 9. d3 {10 +0.2/1 1 288/551/161. OK} a6 {+0.3/27 1128 17070170 1/969/30} 10. a3 {10 +0.2/1 1 276/568/156} Qc7 {+0.2/27 1659 22056318 2/981/17} 11. b4 {10 +0.2/1 1 276/575/149. OK} Nd7 {+0.2/27 2074 30495657 2/983/15. OK} 12. Be3 {10 +0.1/1 1 279/550/171. OK} b6 {+0.2/30 5083 70820589 2/982/16. OK} 13. d4 {10 +0.1/1 1 260/561/179} cxd4 {-0.0/29 3534 46077812 7/988/5} 14. Nxd4 {10 -0.0/1 1 203/559/238} Nxd4 {+0.0/33 2689 39261517 5/990/5} 15. Bxd4 {10 +0.0/1 1 191/611/198} Ne5 {+0.0/36 1568 23900592 5/990/5. OK} 16. Rc1 {10 +0.0/1 1 197/601/202. OK} Bd7 {+0.0/41 1738 25757256 5/990/5. OK} 17. Na4 {10 +0.0/1 1 216/602/182} Bxa4 {+0.0/45 1797 27061071 5/990/5. OK} 18. Qxa4 {10 -0.0/1 1 171/641/188} Nxc4 {+0.0/53 1578 25455191 5/990/5. OK} 19. Bxb6 {10 -0.0/1 1 123/723/154. Best} Nxb6 {+0.0/38 1966 26046102 5/990/5} 20. Rxc7 {10 +0.1/1 1 282/543/175. OK} Nxa4 {+0.0/44 1714 25416361 5/990/5. OK} 21. Rxe7 {10 +0.3/1 1 255/689/56. OK} Nc3 {+0.0/46 1607 23081511 5/990/5. OK} 22. Ra1 {10 +0.1/1 1 158/776/66} a5 {-0.0/49 1974 26270007 5/991/4. OK} 23. bxa5 {10 -0.0/1 1 90/810/100. OK} Rxa5 {-0.0/34 1503 20181836 5/991/4} 24. Rd7 {10 -0.0/1 1 54/878/68. OK} Nb5 {-0.0/36 1471 20167771 5/991/4. OK} 25. a4 {10 +0.0/1 1 53/906/41. OK} g6 {-0.0/38 1491 20334029 5/991/4. OK} 26. Bf1 {10 +0.1/1 1 107/852/41. OK} Nc3 {-0.1/37 1786 25499238 6/991/3. OK} 27. Rxd6 {10 +0.0/1 1 44/913/43. OK} Rfa8 {-0.1/36 1917 28462708 6/991/3. OK} 28. Rd4 {10 +0.0/1 1 44/921/35. OK} Nxa4 {-0.1/35 1813 26668764 6/991/3. OK} 29. Rb4 {10 +0.0/1 1 40/924/36. OK} Nc3 {-0.0/30 1710 25247569 5/991/4} 30. Rxa5 {10 +0.0/1 1 41/923/36. OK} Rxa5 {+0.0/65 1501 21527160 4/992/4} 31. Rc4 {10 -0.0/1 1 20/946/34. OK} Na4 {+0.0/40 1982 24519905 4/992/4. OK} 32. f3 {10 -0.0/1 1 18/950/32. OK} Nc5 {-0.1/34 1458 21015464 7/990/3. OK} 33. h4 {10 -0.0/1 1 17/947/36. OK} h6 {-0.0/33 1616 23157165 5/992/3} 34. Be2 {10 -0.0/1 1 16/952/32} g5 {-0.0/27 1936 25844661 5/991/4} 35. hxg5 {10 -0.0/1 1 11/961/28. OK} hxg5 {+0.0/41 1525 20167917 4/992/4} 36. f4 {10 -0.0/1 1 10/970/20. OK} Nd7 {+0.0/51 1572 20552752 4/992/4} 37. Kf2 {10 -0.0/1 1 8/976/16. OK} gxf4 {+0.0/43 1588 19272106 4/992/4} 38. gxf4 {10 +0.0/1 1 4/990/6} Kf8 {+0.0/49 1523 19429498 4/992/4} 39. e5 {10 +0.0/1 1 6/986/8. OK} Nb6 {+0.0/60 1534 21060273 4/992/4} 40. Rd4 {10 +0.0/1 1 5/987/8. OK} Rc5 {+0.0/53 1509 19343976 4/992/4} 41. Bf3 {10 +0.0/1 1 5/988/7. OK} Rc2+ {+0.0/51 1565 20993389 4/992/4} 42. Kg3 {10 +0.0/1 1 7/985/8. OK} Kg7 {+0.0/69 1481 19281954 4/992/4} 43. Rd6 {10 +0.0/1 1 5/988/7. OK} Nc4 {+0.0/57 1531 19948467 4/992/4} 44. Rc6 {10 +0.0/1 1 15/978/7. OK} Ne3 {+0.0/71 1521 23097786 5/990/5. OK} 45. Rxc2 {10 +0.0/1 1 6/991/3. OK} Nxc2 {+0.0/79 1651 28822594 5/990/5} 46. f5 {10 +0.0/1 1 3/994/3} exf5 {adjudication by egtb +0.0/183 1977 37108983 5/990/5. Loses game: --- --- Missed win: --- --- Mistake: --- --- Inaccurate: --- --- OK: White=28 Black=18 Best: White=1 --- Strong: --- --- Brilliant: --- --- . Weighted Error Value: White=0.01/Black=0.00 Centipawn loss: w=2/b=1} 1/2-1/2

[/pgn]

[pgn][Event "banksia game"]
[Site "*"]
[Date "2023.01.06"]
[Round "1"]
[Board "47"]
[White "Lc0 v0.29.0"]
[Black "Stockfish dev-20230104"]
[Result "1/2-1/2"]
[ECO "B10"]
[PlyCount "52"]
[EventDate "2023.??.??"]

1. e4 c6 2. Nc3 d5 3. Nf3 dxe4 {B10: Caro-Kann, two knights variation} 4. Nxe4 Nf6 {Stockfish dev-20230104-9fe9ff00:} 5. Qe2 Nxe4 6. Qxe4 Nd7 {End of opening} 7. Bc4 {+0.2/1 2450 1 281/582/137. OK} Nf6 {+0.5/30 4427 59728433 1/921/78. OK} 8. Ne5 {10 +0.2/1 1 290/559/151. OK} e6 {+0.4/27 1532 23541909 1/945/54. OK} 9. Qe2 {10 +0.2/1 1 298/547/155. OK} b5 {+0.4/30 1691 25673809 1/944/55. OK} 10. Bb3 {10 +0.3/1 1 354/488/158. Inaccurate} Qc7 {+0.3/29 2955 41441954 2/975/23. OK} 11. d4 {10 +0.2/1 1 286/559/155} Bd6 {+0.2/31 1543 23504014 2/978/20. OK} 12. O-O {10 +0.2/1 1 286/565/149} O-O {+0.1/31 3571 53900590 4/987/9. OK} 13. Re1 {10 +0.2/1 1 272/595/133} a5 {+0.0/32 1802 26786400 6/988/6. OK} 14. c3 {10 +0.2/1 1 268/588/144. OK} c5 {+0.1/38 2821 42341698 3/988/9. OK} 15. Bg5 {10 +0.2/1 1 266/590/144} Bb7 {+0.0/32 1759 24996922 5/990/5. OK} 16. Bxf6 {10 +0.3/1 1 353/512/135. OK} gxf6 {+0.0/42 2671 35639532 5/990/5} 17. Qh5 {10 +0.2/1 1 318/534/148. OK} fxe5 {+0.0/48 1899 26444410 5/990/5. OK} 18. Qg5+ {10 +0.1/1 1 133/821/46. OK} Kh8 {+0.0/52 1595 21346514 5/990/5} 19. Qf6+ {10 +0.3/1 1 204/775/21. Best (Easy)} Kg8 {+0.0/56 571 8470418 5/990/5} 20. Qg5+ {10 +0.3/1 1 268/697/35. OK} Kh8 {+0.0/55 1216 16803558 5/990/5} 21. Qf6+ {10 +0.0/1 1 52/930/18. Best (Easy)} Kg8 {+0.0/64 556 7853620 5/990/5} 22. Re3 {10 +0.1/1 1 143/812/45. OK} exd4 {+0.0/56 2879 37115637 5/990/5} 23. Qg5+ {10 -0.0/1 1 17/958/25. OK} Kh8 {+0.0/93 1631 23022414 5/991/4} 24. Qf6+ {10 +0.0/1 1 20/971/9. Best (Easy)} Kg8 {+0.0/63 614 8934242 5/991/4} 25. Qg5+ {10 -0.0/1 1 7/976/17} Kh8 {+0.0/67 972 12434561 4/992/4} 26. Qf6+ {10 +0.0/1 1 11/980/9. Best (Easy)} Kg8 {repetition +0.0/245 24 3920 4/992/4. Loses game: --- --- Missed win: --- --- Mistake: --- --- Inaccurate: White=1 --- OK: White=10 Black=10 Best: White=4 --- Strong: --- --- Brilliant: --- --- . Weighted Error Value: White=0.04/Black=0.01 Centipawn loss: w=4/b=1} 1/2-1/2

[/pgn]

Posts : 3110 Join date : 2020-11-18

I strongly approve of this testing: IMO "knowing" is more intelligent than "calculating".

Posts : 880 Join date : 2020-11-25 Location : USA

TheSelfImprover wrote:: I strongly approve of this testing: IMO "knowing" is more intelligent than "calculating".

I like learning new things about chess and chess engines.. And to do this you have to push the chess engines. And do crazy types of testing.

I never would have guessed you could play chess well enough to draw Stockfish with zero calculation. Or play chess at this type of strength. With just positional understanding.

I am not buying it totally. As if you know your opponent is not calculating. You would change your playing style. But clearly Lc0 can play chess at some type of master level. With zero look ahead and just with positional understanding.