Crowd Sourced processing of Super GMs + Hans for engine correlation

[adnoh]

I'm running it with Komodo 13.3 right now. I have the latest Komodo's if there is value to having analysis from any specific version.

Code: Select all

Microsoft Windows [Version 10.0.22000.1042]
(c) Microsoft Corporation. All rights reserved.

C:\>cd Users\Charles\Downloads\CC-Analysis

C:\Users\Charles\Downloads\CC-Analysis>C:\Users\Charles\AppData\Local\Programs\Python\Python310\python.exe process_games.py
Finished analysis for Carlsen/1990223.pgn in 147.74 seconds
Finished analysis for Carlsen/1990199.pgn in 150.63 seconds
Finished analysis for Carlsen/1993640.pgn in 170.15 seconds
Finished analysis for Carlsen/1990906.pgn in 192.96 seconds
Finished analysis for Carlsen/1994008.pgn in 198.39 seconds
Finished analysis for Carlsen/1990389.pgn in 204.80 seconds
Finished analysis for Carlsen/1990937.pgn in 206.40 seconds
Finished analysis for Carlsen/1994006.pgn in 206.83 seconds
Finished analysis for Carlsen/1990437.pgn in 208.65 seconds
Finished analysis for Carlsen/1990509.pgn in 215.90 seconds
Finished analysis for Carlsen/1993656.pgn in 227.75 seconds

[dkappe]

Critter 1.6 and other older engines.

[AndrewGrant]

Data uploaded for Ethereal 13.00, and 13.25. Now clue how anyone wants to "process" said data, but the T1/T2/T3 metric that people enjoyed so much is here: Not sure how interesting or useful it is though.

Code: Select all

      Ethereal-13.00              Carlsen: T1=0.572 T2=0.186 T3=0.094 TX=0.852
      Ethereal-13.00             Firouzja: T1=0.547 T2=0.189 T3=0.101 TX=0.836
      Ethereal-13.00                 Giri: T1=0.541 T2=0.192 T3=0.101 TX=0.833
      Ethereal-13.00                Liren: T1=0.558 T2=0.197 T3=0.092 TX=0.847
      Ethereal-13.00             Nakamura: T1=0.550 T2=0.187 T3=0.098 TX=0.836
      Ethereal-13.00       Nepomniachtchi: T1=0.559 T2=0.191 T3=0.094 TX=0.843
      Ethereal-13.00              Niemann: T1=0.560 T2=0.193 T3=0.092 TX=0.845
      Ethereal-13.00                   So: T1=0.577 T2=0.188 T3=0.097 TX=0.863

Code: Select all

      Ethereal-13.25              Carlsen: T1=0.570 T2=0.189 T3=0.093 TX=0.852
      Ethereal-13.25             Firouzja: T1=0.550 T2=0.190 T3=0.095 TX=0.835
      Ethereal-13.25                 Giri: T1=0.546 T2=0.195 T3=0.095 TX=0.836
      Ethereal-13.25                Liren: T1=0.560 T2=0.192 T3=0.094 TX=0.846
      Ethereal-13.25             Nakamura: T1=0.554 T2=0.191 T3=0.097 TX=0.841
      Ethereal-13.25       Nepomniachtchi: T1=0.567 T2=0.184 T3=0.094 TX=0.845
      Ethereal-13.25              Niemann: T1=0.565 T2=0.187 T3=0.088 TX=0.841
      Ethereal-13.25                   So: T1=0.582 T2=0.187 T3=0.092 TX=0.861

[jhellis3]

I mentioned this in the SF discord... An interesting metric might be the % of "singular" nodes a player encounters in their games (those where the best move is clearly better than all the alternatives).

[dkappe]

I mentioned this in the SF discord... An interesting metric might be the % of "singular" nodes a player encounters in their games (those where the best move is clearly better than all the alternatives).

See my “Chess Nose.”

Here the white side of this game.

[pgn] [Event "Casual Blitz game"]
[Site "https://lichess.org/N5O66Ode"]
[Date "2020.05.10"]
[White "dkappe"]
[Black "ShMat"]
[Result "1-0"]
[UTCDate "2020.05.10"]
[UTCTime "14:57:43"]
[WhiteElo "2114"]
[BlackElo "2649"]
[BlackTitle "IM"]
[Variant "Standard"]
[TimeControl "180+2"]
[ECO "B01"]
[Opening "Scandinavian Defense: Mieses-Kotroc Variation"]
[Termination "Normal"]
[Annotator "lichess.org"]

1. e4 d5 2. exd5 Qxd5 { B01 Scandinavian Defense: Mieses-Kotroc Variation } 3. Nc3 Qd8 4. d4 Nf6 5. f4?! { (0.50 → -0.23) Inaccuracy. Nf3 was best. } (5. Nf3 e6 6. Bc4 a6 7. a4 Bb4 8. O-O b6 9. d5 Bxc3) 5... e6 6. Nf3 Bb4 7. Bd3 O-O 8. O-O Nc6 9. Be3 b6 10. a3?! { (0.56 → 0.00) Inaccuracy. Qe2 was best. } (10. Qe2 Bxc3 11. bxc3 Ne7 12. Ne5 Bb7 13. a4 Nf5 14. a5 Be4 15. Bc1 Bxd3 16. cxd3 h5) 10... Bxc3 11. bxc3 Nd5 12. Bd2 Qd6 13. Ne5 Bb7 14. Qh5?! { (0.95 → 0.17) Inaccuracy. Nc4 was best. } (14. Nc4 Qd8 15. f5 exf5 16. Bxf5 g6 17. Bd3 Qd7 18. Qf3 f5 19. Rae1 Rae8 20. Ne3 Na5) 14... g6?! { (0.17 →
1.00) Inaccuracy. f5 was best. } (14... f5 15. Nc4 Qd7 16. Rfe1 Nf6 17. Qe2 Rae8 18. h3 a6 19. a4 h6 20. Qf2 Qd5 21. Qg3) 15. Qh6 Nde7?? { (0.90 → 6.38) Blunder. f5 was best. } (15... f5) 16. Rf3 Nf5?! { (5.61 → 11.36) Inaccuracy. f5 was best. } (16... f5 17. Rh3) 17. Bxf5 exf5?! { (11.25 → Mate in 4) Checkmate is now unavoidable. Nxe5 was best. } (17... Nxe5 18. Rh3 Rfd8 19. fxe5 Qc6 20. Qxh7+ Kf8 21. Be4 Qxe4 22. Rf3 Qxf3 23. gxf3 Bd5 24. Bg5) 18. Rh3 Rfd8 19. Qxh7+ Kf8 20. Qxf7# { White wins by checkmate. } 1-0[/pgn]

A “good” move is within 100 cp of the best move.

ShMat:
10 total positions, 2 (20.00%) total matches, 1 goof(s), 3 blunder(s)
one good move: 1 (10.00%), matched 0 (0.00%)
2-3 good moves: 3 (30.00%), matched 1 (33.33%)
4+ good moves: 6 (60.00%), matched 1 (16.67%)

I suspect the white player is a cheater.

[jhellis3]

I am not sure you discerned the point....

[dkappe]

I am not sure you discerned the point....

If we take a less silly example:

[pgn] [Event "Julius Baer Generation Cup"]
[Site "chess24.com INT"]
[Date "2022.09.19"]
[EventDate "2022.09.18"]
[Round "7.6"]
[Result "1-0"]
[White "Magnus Carlsen"]
[Black "Levon Aronian"]
[ECO "E04"]
[WhiteElo "2861"]
[BlackElo "2759"]
[PlyCount "106"]

1. Nf3 d5 2. g3 c5 3. Bg2 Nc6 4. d4 e6 5. O-O Nf6 6. c4 dxc4 7. Qa4 Bd7 8. Qxc4
b5 9. Qd3 c4 10. Qd1 Rc8 11. Bf4 h6 12. Nc3 b4 13. Nb5 Qa5 14. Nd6+ Bxd6 15.
Bxd6 Ne4 16. Bf4 c3 17. Qb1 f5 18. a3 O-O 19. axb4 Qb5 20. Re1 a5 21. bxa5 g5
22. bxc3 Qxb1 23. Rexb1 gxf4 24. Rb7 Be8 25. gxf4 Nxc3 26. e3 Nd5 27. a6 Rf7 28.
Ne5 Nxe5 29. Bxd5 exd5 30. fxe5 Ra8 31. Rxf7 Kxf7 32. a7 Bb5 33. Kg2 f4 34. exf4
Kg6 35. Kf3 Kf5 36. Rb1 Ba4 37. Rb7 Bd1+ 38. Ke3 Bc2 39. f3 Ke6 40. Kd2 Bf5 41.
Kc3 Bh3 42. Kb4 Kf5 43. Rf7+ Kg6 44. Rf6+ Kh5 45. Ra6 Bc8 46. Ra2 Kg6 47. Kc5
Kf5 48. Kb6 Kxf4 49. Rc2 Bf5 50. Rc6 Ke3 51. Kb7 Re8 52. a8=Q Rxa8 53. Kxa8 1-0[/pgn]

Carlsen:
45 total positions, 32 (71.11%) total matches, 1 goof(s), 0 blunder(s)
one good move: 14 (31.11%), matched 13 (92.86%)
2-3 good moves: 8 (17.78%), matched 7 (87.50%)
4+ good moves: 23 (51.11%), matched 12 (52.17%)

Aronian:
44 total positions, 30 (68.18%) total matches, 4 goof(s), 0 blunder(s)
one good move: 11 (25.00%), matched 11 (100.00%)
2-3 good moves: 16 (36.36%), matched 10 (62.50%)
4+ good moves: 17 (38.64%), matched 9 (52.94%)

The “one good move” positions correspond to your singular nodes. As expected, the positions with many “good” moves have a lower match %.

[jhellis3]

I am not sure you discerned the point....

[dkappe]

I am not sure you discerned the point....

[jhellis3]

You want a hint?

The match % is irrelevant. Whether they even played the singular move or not is irrelevant. Also, cherry picked examples are..... you guessed it, irrelevant.

If that doesn't provide enough to make the proper discernment, well...