Komodo Dragon vs. GM Perelshteyn, knight odds match.

[Chessqueen]

Where can I watch the Match ? Direct Link ?

I guess that Komodo Dragon can only beat a GM rated over 2500 with TC of 10'+5". Is this the MCTS or standard Komodo Dragon playing and how many threads ? Ask him to play 2 extra games at TC of 10'+5"

[Chessqueen]

Where can I watch the Match ? Direct Link ?

I guess that Komodo Dragon can only beat a GM rated over 2500 with TC of 10'+5". Is this the MCTS or standard Komodo Dragon playing and how many threads

Actually Komodo Dragon performed very good for him having the latest commercial version to practice against, next time play at this TC of 10'+5"

[lkaufman]

So the final score was 4 to 2 in favor of GM Perelshteyn, exactly the same ratio as in our previous match with the similarly rated GM Wirig (we lost 6 to 3, same conditions), and half a point worse for Dragon than my prediction. This time we won one (the last game), lost three, and drew two. One difference from the previous match was that Komodo played as if it were blitz (5 sec/move average). We used MCTS for the first three games (using all but one of the 32 cores), but only scored half a point, and lost the third game quite badly. So at the halfway break I switched to standard mode, and we did better (one draw, one loss, one win). Of course the sample size is too small to say anything with confidence, but I felt that Dragon played better in standard mode under these conditions. Perelshteyn played quite well, making no obvious blunders; in all previous knight odds matches, the human GM or IM or FM made at least one obvious blunder. Dragon at least avoided losing the majority of the games, making this the third knight odds match (including IM Toth) in a row in which Dragon has accomplished this feat. But it is apparently not yet strong enough to win or even draw a knight odds Rapid match against a 2500 GM; it can often save draws in bad endgames, but winning is very difficult at knight odds.
I also note that Dragon scored only 0.5 out of 3 with b1 knight odds, but 1.5 out of 3 with g1 knight odds. Here too the sample size is too tiny to say much, but my subjective opinion is that Dragon gets better positions with g1 odds than with b1 odds, primarily because 1.e4 and 1.c4 are both good move at g1 odds but poor moves at b1 odds. For the future, we could just do g1 odds (it is the odds-giver's choice normally of which knight to remove), but there is something to be said for variety. Or we could change the ratio of g1 to b1 odds games.
Games can be viewed or downloaded by going to chess.com archives and looking at the latest games of PlayKomodo (or EugenePerelshteyn).
I remain convinced that the goal of defeating a 2500 FIDE GM in a Rapid knight odds match is a realistic goal for the next year, but we'll need to find ways to pose more problems than we do now.

[Chessqueen]

Here is the 6th game as it was played up to the 7th move[/size] This is a good way to find out if the human GM indeed plays better than Gaviota 0.84 win 32 with Knight Odds and the other 5 games can also be replaced with Gaviota 0.84 win 32 after the first 7 opening moves, by doing this we can figure out if the opening choice of moves was what caused Dragon 2.5.1.x to lose the Match, or is it that the GM is much stronger than Gaviota 0.84 win 32. I suppose that selecting very aggressive Opening scare human GMs, and therefore, the contempt should be lower a little bit to make it more aggressive, remember if all the pieces and pawns are exchanged with the extra knight the human can only get a draw

[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.11.06"]
[Round "6"]
[White "Dragon-2.5-64bit-avx2"]
[Black "Gaviota-win32-0.84"]
[Result "1-0"]
[BlackElo "2500"]
[Time "15:59:34"]
[WhiteElo "2500"]
[TimeControl "180+2"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "61"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e5 2. Nc3 Nf6 3. d4 exd4 4. Qxd4 Nc6 5. Qe3 Bb4 6. Bd2 O-O 7. O-O-O
Re8 8. Bc4 Bxc3 9. Bxc3 Nxe4 10. Rhe1 h6 11. Qf4 Nd6 12. Rxd6 Rxe1+ 13.
Bxe1 cxd6 14. Qxf7+ Kh7 15. f4 d5 16. Bxd5 Qe7 17. Qg8+ Kg6 18. Bh4 Qe3+
19. Kd1 Qg1+ 20. Ke2 Nd4+ 21. Kd2 Nf3+ 22. gxf3 Qd4+ 23. Ke2 Qxd5 24. Qxd5
Kh7 25. Qf7 Kh8 26. Bf2 b6 27. Bd4 Ba6+ 28. Kf2 Rg8 29. Qg6 Bd3 30. cxd3 d5
31. Qxh6# *[/pgn]

[Chessqueen]

Here is the 6th game as it was played up to the 7th move[/size] This is a good way to find out if the human GM indeed plays better than Gaviota 0.84 win 32 with Knight Odds and the other 5 games can also be replaced with Gaviota 0.84 win 32 after the first 7 opening moves, by doing this we can figure out if the opening choice of moves was what caused Dragon 2.5.1.x to lose the Match, or is it that the GM is much stronger than Gaviota 0.84 win 32. I suppose that selecting very aggressive Opening scare human GMs, and therefore, the contempt should be lower a little bit to make it more aggressive, remember if all the pieces and pawns are exchanged with the extra knight the human can only get a draw

Note: I just finished with my little experiment the conclusion is that a GM like Perelsteyn rated 2509 play better than any engine rated similar by CCRL at Knight Odds Vs Komodo Dragon regardless of the contempt settings either =100 or 75

[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.11.06"]
[Round "6"]
[White "Dragon-2.5-64bit-avx2"]
[Black "Gaviota-win32-0.84"]
[Result "1-0"]
[BlackElo "2500"]
[Time "15:59:34"]
[WhiteElo "2500"]
[TimeControl "180+2"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "61"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e5 2. Nc3 Nf6 3. d4 exd4 4. Qxd4 Nc6 5. Qe3 Bb4 6. Bd2 O-O 7. O-O-O
Re8 8. Bc4 Bxc3 9. Bxc3 Nxe4 10. Rhe1 h6 11. Qf4 Nd6 12. Rxd6 Rxe1+ 13.
Bxe1 cxd6 14. Qxf7+ Kh7 15. f4 d5 16. Bxd5 Qe7 17. Qg8+ Kg6 18. Bh4 Qe3+
19. Kd1 Qg1+ 20. Ke2 Nd4+ 21. Kd2 Nf3+ 22. gxf3 Qd4+ 23. Ke2 Qxd5 24. Qxd5
Kh7 25. Qf7 Kh8 26. Bf2 b6 27. Bd4 Ba6+ 28. Kf2 Rg8 29. Qg6 Bd3 30. cxd3 d5
31. Qxh6# *[/pgn]

[Chessqueen]

Experiment #3 For this game I took one of the game played yesterday up to move 7. 0-0
Bg4. I set it up to Contempt = 100 much defensive than for contempt = 65. You will notice that move 11. h3 is one of the defensive move made here. I used the standard mode NOT MCTS.


[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.11.07"]
[Round "?"]
[White "Dragon-2.5-64bit-avx2"]
[Black "Gaviota-win32-0.84"]
[Result "0-1"]
[BlackElo "2500"]
[Time "05:29:47"]
[WhiteElo "2500"]
[TimeControl "180+2"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "107"]
[WhiteType "program"]
[BlackType "program"]

1. Nf3 d5 2. e3 Nf6 3. b3 g6 4. c4 c6 5. Bb2 Bg7 6. Be2 O-O 7. O-O Bg4 8.
d4 Qa5 9. Qc2 Bf5 10. Qc1 Ne4 11. h3 Nd7 12. a4 dxc4 13. bxc4 e5 14. Re1
Rfe8 15. Bf1 h5 16. c5 exd4 17. exd4 Ndf6 18. Ra3 Nd5 19. Bc4 Kh7 20. Qa1
Bh6 21. Ba2 Nd2 22. Ne5 f6 23. Bxd5 cxd5 24. Bc3 Qc7 25. Rd1 Ne4 26. Nd3
Nxc3 27. Qxc3 Re2 28. Nb4 Qf4 29. Qf3 Rae8 30. Rf1 Qxd4 31. Nxd5 Kg7 32.
Raa1 Rd2 33. Nc3 Qxc5 34. Rad1 Qc6 35. Qxc6 bxc6 36. Nb1 Rb2 37. a5 Re5 38.
Na3 Rb4 39. a6 Rd5 40. Ra1 Ra4 41. Rae1 Rxa3 42. Re7+ Kf8 43. Rxa7 Bd3 44.
Ra8+ Kf7 45. Ra7+ Ke6 46. Rd1 Rxa6 47. Rc7 Be2 48. Re1 Rd1 49. Rxd1 Bxd1
50. Rc8 Ra1 51. Re8+ Kf5 52. Re1 Bd2 53. Rf1 c5 54. Kh2 {White resigns} *[/pgn]

[Chessqueen]

Experiment #4 For this game I took one of the game played yesterday up to move 7. 0-0
Bg4 I set it up to Contempt = 65 much aggressive than for contempt = 100. You will notice that 11. Nh4 is one of the aggressive move made here. I used the standard mode NOT MCTS. I read long time ago that Mike Tyson trainer once told him when he started boxing you do not win a fight by holding back and wait for your opponent to start attacking you with punches, you have to be the first to scare your opponent with your punches and beat him to death.


[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2021.11.07"]
[Round "?"]
[White "Dragon-2.5-64bit-avx2"]
[Black "Gaviota-win32-0.84"]
[Result "1-0"]
[BlackElo "2500"]
[Time "05:29:47"]
[WhiteElo "2500"]
[TimeControl "180+2"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[Termination "normal"]
[PlyCount "125"]
[WhiteType "human"]
[BlackType "human"]

1. Nf3 d5 2. e3 Nf6 3. b3 g6 4. c4 c6 5. Bb2 Bg7 6. Be2 O-O 7. O-O Bg4 8.
d4 Qa5 9. Qc2 Bf5 10. Qc1 Ne4 11. Nh4 Be6 12. Rd1 Na6 13. Qc2 Bf6 14. Nf3
Nb4 15. Qc1 dxc4 16. bxc4 Qa4 17. Ne1 b6 18. a3 Na6 19. Nf3 Rab8 20. h3 Nc7
21. Bd3 Nd6 22. Nd2 b5 23. c5 Nc4 24. Bc2 Qa5 25. Nb3 Qa6 26. Bc3 Bg5 27.
Re1 Nd5 28. Bd2 h6 29. Kh2 Bf5 30. Bd1 h5 31. f4 Nxd2 32. Qxd2 Bh6 33. g4
hxg4 34. hxg4 Be4 35. Qf2 Nc3 36. Nd2 Nxd1 37. Rexd1 Bd5 38. e4 Bc4 39. Kg3
Bg7 40. Nf3 Rfd8 41. f5 Qa4 42. Kg2 e5 43. Rd2 Qa5 44. Qh4 Qc3 45. Rh1 gxf5
46. gxf5 f6 47. Rf2 exd4 48. Rh3 d3 49. Kh2 Kf7 50. Rg2 Rg8 51. Qg3 Rb7 52.
Rh7 Ke8 53. Qd6 Re7 54. e5 Rd7 55. Qb8+ Rd8 56. Qb7 Rd7 57. Qc8+ Rd8 58.
Qxc6+ Rd7 59. e6 Bxe6 60. fxe6 Kf8 61. Qc8+ Ke7 62. Qxd7+ Kf8 63. Qd8# *[/pgn]