Deep Blue - Kasparov, 1997 game 2

[Dann Corbit]

Here is some analysis I did against game 2 positions, some time ago...

[d]1b4k1/6p1/5p1p/Pp1PpP2/2p5/2P4P/2B3P1/6K1 w - - acd 33; acn 39024361821; acs 10720804; bm Be4; ce 350; id "NM(pv)-1j"; pv Be4 Ba7+ Kf1 Kf7 Ke2;

[d]r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - acd 32/87; acn 27740650800; acs 1537202; bm Be4; ce 82; pv Be4 Rcb8 Ra6 Qd8 Kh2 Kh7 g4 Rxa6 Rxa6 Rb7 Kg2 Kg8 Qa2 Kf7 Kg3 Ke7 Qa3 Qc7 Qa1 Kf7 Qg1 Qd8 Kg2 Kg8 Qf2 Kh8 Qa2 Kh7 Kf3 Bb8 Ra8 Qc7 d6 Qd7 Bxb7 Qxb7+ Kg3 Bxd6;

[d]r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - acd 30; acn 4505215255; acs 1123243; bm Be4; ce 60; pv Be4 Qd8 Kh2 Rxa2 Qxa2 Rb8 g3 Kf8 Kg2 Bc7 Qa7 Bb6 Qa6 Bc7 Qc6 Rb6 Qa8;

[d]r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - acd 26/77; acn 6379248338; acs 388854; bm axb5; ce 80; pv axb5 axb5 Be4 Qd8 Kh2 Rcb8 Ra6 Rxa6 Rxa6 Rb7 Qa2 Kh7 Qe2 Kg8 Qf2 Kh8 Qe3 Kg8 g4 Kh7 Qf2 Kh8 Qa2 Rb8 Ra7 Qb6 Kh1 Rd8;

I really don't have the experience, or the software, to decipher this. Is this convincing evidence that a decent program would not allow ...Qe3? Can it be rephrased in chessic terms? My chess is a lot better than my code reading.

It is more likely that simply programs don't understand what is going on. And whether they play Kh1 or something else is based on random luck rather than understanding here.

This draw is _deep_. Several of us spent all night using every computer we had to convince everyone it was really drawn. Until then, even the best player on the planet thought it was lost.

Rybka isn't God, but I would like to point out a few things here:
[d]1b4k1/6p1/5p1p/Pp1PpP2/2p5/2P4P/2B3P1/6K1 w - - acd 33; acn 39024361821; acs 10720804; bm Be4; ce 350; id "NM(pv)-1j"; pv Be4 Ba7+ Kf1 Kf7 Ke2;

The depth is 33 plies, and Rybka under-reports depth. The analysis ran for 50 hours.

Here is my original log for that position:

Code: Select all

Analysis from C:\tmp\dbfin.epd   
10/25/2008 12:43:25 AM Level: 180000 Seconds
Analyzing engine: Rybka 3

1) a6 Kb6 Kh5 Kxa6 Kg6  
    Avoid move: 
    Best move (Rybka 3): Bc2-e4
    Not found in: 50:00:00
      2 00:08            224         1.835   +0.61  Kg1f2
      4 00:08            280         448      +0.44  a5a6 Kg8f8
      5 00:08            288         449      +0.29  a5a6 Kg8f8
      5 00:08            295         411      +0.31  Kg1f1 Kg8f7
      6 00:08            298         415      +0.31  Kg1f1 Kg8f7
      7 00:08            412         574      +0.30  Kg1f1 Kg8f7
      8 00:08            458         545      +0.29  Kg1f1 Kg8f7
      9 00:08            690         793      +0.30  Kg1f1 Kg8f7
     10            00:08          6.788        7.418   +0.29  a5a6 Kg8f8
     11            00:08          8.295        9.065   +0.29  a5a6 Kg8f8 Kg1f2 Kf8e7
     12            00:08         12.073       12.771 +0.29  a5a6 Kg8f8 Kg1f2 Kf8e7 Bc2e4 Ke7d6 Kf2g3 Bb8a7 Kg3g4
     13+         00:09         30.468       29.405 +0.49  a5a6
     13            00:09         35.113       33.385 +0.29  a5a6 Kg8f8 Kg1f2 Kf8e7 Bc2e4 Ke7d6 Kf2g3 Kd6c5 Kg3g4
     14+         00:09         60.599       50.988 +0.49  a5a6
     14+         00:09        118.652      70.762 +0.69  a5a6
     14            00:09        144.943      77.343 +0.77  a5a6 Kg8f8 Kg1f2
     15            00:09        153.743      78.208 +0.77  a5a6 Kg8f8 Kg1f2
     16            00:12        744.360      153.612           +0.94  a5a6 e5e4 Bc2xe4 Kg8f7 d5d6 Bb8xd6 Kg1f1 Bd6b8 Be4c6 Kf7e7 Bc6xb5 Ke7d6 Bb5xc4 Kd6c5 Bc4e2 Kc5b6 Kf1e1 Kb6a7 Ke1d2 Bb8e5 Be2b5
     16+         00:12        747.716      153.839           +0.97  Kg1f2
     16            00:13        791.846      152.846           +1.08  Kg1f2 Kg8f7 Bc2e4 Kf7e7 Kf2f3 Ke7d6 Kf3g4 Kd6c5 a5a6
     17            00:14      1.007.807     158.232           +0.77  a5a6 Kg8f8 Kg1f2
     18            00:15      1.074.114     155.286           +0.80  a5a6 Kg8f8 Kg1f2 e5e4 Bc2xe4 Kf8e7 Kf2f3 Ke7d6 Kf3g4 Kd6c7 Kg4h5 Kc7b6 Kh5g6 Kb6xa6 Kg6xg7 Bb8e5 d5d6 Ka6b6 d6d7 Kb6c7 Be4c6 Be5xc3 Bc6xb5 Bc3e5 Bb5xc4 Kc7xd7 Kg7xh6 Be5g3 Kh6g6 Kd7e7 Kg6h5
     19            00:16      1.272.699     155.555           +0.82  a5a6 Kg8f8 Kg1f2 e5e4 Bc2xe4 Kf8e7 Kf2f3 Ke7d6 Kf3g4 Kd6c7 Kg4h5 Kc7b6 Kh5g6 Kb6xa6 Kg6xg7 Bb8e5 d5d6 Ka6b6 d6d7 Kb6c7 Be4c6 Be5xc3 Bc6xb5 Bc3e5 Bb5xc4 Kc7xd7 Kg7xh6 Be5g3 Kh6g6 Kd7e7 Bc4b3
     20            00:21      2.085.918     161.266           +1.01  a5a6 e5e4 Bc2xe4 Kg8f7 d5d6 Bb8xd6 Kg1f1 Bd6b8 Be4c6 Kf7e7 Bc6xb5 Ke7d6 Bb5xc4 Kd6c5 Bc4e2 Kc5b6 Kf1e1 h6h5 Ke1d2 Bb8a7 Kd2c2 Ba7b8 Kc2b3 Bb8e5 Kb3b4
     20+         00:29      3.084.117     149.398           +1.02  Kg1f2 Kg8f7
     20            00:38      4.155.743     138.398           +1.09  Kg1f2 Kg8f7
     21            01:28     12.974.874    165.501           +1.17  Kg1f2 e5e4 Kf2e3 Kg8f7 d5d6 Bb8xd6 a5a6 b5b4 Ke3e2 Bd6b8 c3xb4 Kf7e7 Bc2xe4 Ke7d6 Ke2d2 Kd6e5 Be4c6 Ke5d4 Kd2c2 Bb8a7 g2g3 Ba7b6 h3h4 Bb6a7 g3g4 Ba7b8 Kc2d2 Bb8f4+ Kd2e2
     21+         01:28     13.021.740    165.745           +1.42  Bc2e4 Kg8f7 Kg1h2 Kf7e7 a5a6 Ke7d6 Kh2g3 Kd6c7 Kg3h4 Kc7b6 Kh4h5 Bb8c7 Kh5g6 Bc7d8 Kg6xg7 Kb6xa6 Kg7f7 Ka6b6 Kf7e8 Bd8c7 Ke8d7 Bc7b8 Kd7e6 Kb6c5 Ke6xf6
     21            01:28    13.021.748    165.713           +1.42  Bc2e4 Kg8f7 Kg1h2 Kf7e7 a5a6 Ke7d6 Kh2g3 Kd6c7 Kg3h4 Kc7b6 Kh4h5 Bb8c7 Kh5g6 Bc7d8 Kg6xg7 Kb6xa6 Kg7f7 Ka6b6 Kf7e8 Bd8c7 Ke8d7 Bc7b8 Kd7e6 Kb6c5 Ke6xf6
     22            01:28     13.021.762    165.713           +1.42  Bc2e4 Kg8f7 Kg1h2 Kf7e7 a5a6 Ke7d6 Kh2g3 Kd6c7 Kg3h4 Kc7b6 Kh4h5 Bb8c7 Kh5g6 Bc7d8 Kg6xg7 Kb6xa6 Kg7f7 Ka6b6 Kf7e8 Bd8c7 Ke8d7 Bc7b8 Kd7e6 Kb6c5 Ke6xf6
     23            01:28     13.021.776    165.680           +1.42  Bc2e4 Kg8f7 Kg1h2 Kf7e7 a5a6 Ke7d6 Kh2g3 Kd6c7 Kg3h4 Kc7b6 Kh4h5 Bb8c7 Kh5g6 Bc7d8 Kg6xg7 Kb6xa6 Kg7f7 Ka6b6 Kf7e8 Bd8c7 Ke8d7 Bc7b8 Kd7e6 Kb6c5 Ke6xf6
     24            03:09     29.764.709    167.805           +1.54  Bc2e4 Bb8a7+ Kg1f1 Kg8f7 d5d6 Ba7c5
     25            04:23     42.565.380    170.657           +1.54  Bc2e4 Bb8a7+ Kg1f1 Kg8f7 d5d6 Ba7c5 a5a6 Kf7e8 Be4c6+ Ke8d8 Bc6xb5 Kd8c8 d6d7+ Kc8d8 Bb5c6 e5e4 Kf1e2 Bc5b6 Bc6a4 h6h5 h3h4 Bb6a7 Ba4b5
     26            06:42     67.530.076    175.164           +1.54  Bc2e4 Bb8a7+ Kg1f1 Kg8f7 d5d6 Ba7c5 a5a6 Kf7e8 Be4c6+ Ke8d8 Bc6xb5 Kd8c8 d6d7+ Kc8d8 Bb5c6 e5e4 Kf1e2 Bc5b6 Bc6a4 h6h5 h3h4 Bb6a7 Ba4b5
     27            09:17     98.736.128    183.912           +1.54  Bc2e4 Bb8a7+ Kg1f1 Kg8f7 d5d6 Ba7c5 a5a6 Kf7e8 Be4c6+ Ke8d8 Bc6xb5 Kd8c8 d6d7+ Kc8d8 Bb5c6 e5e4 Kf1e2 Bc5b6 Bc6a4 h6h5 h3h4 Bb6a7 Ba4b5
     28            12:28   137.325.830   189.995           +1.54  Bc2e4 Bb8a7+ Kg1f1 Kg8f7 d5d6 Ba7c5 a5a6 Kf7e8 Be4c6+ Ke8d8 Bc6xb5 Kd8c8 d6d7+ Kc8d8 Bb5c6 e5e4 Kf1e2 Bc5b6 Bc6a4 h6h5 h3h4 Bb6a7 Ba4b5
     29+         29:55   328.150.541   187.992           +1.74  Bc2e4
     29            57:28   655.237.003   194.989           +1.70  Bc2e4 b5b4 c3xb4 Bb8d6 b4b5 Kg8f8
     30+         1:24:07            992.127.838   201.590           +1.90  Bc2e4
     30+         2:53:35            2.120.558.117 208.645           +2.10  Bc2e4
     30+         13:02:41          9.681.301,241 211.139           +3.30  Bc2e4
     30            13:04:53          9.709.721,091 211.165           +3.30  Bc2e4 b5b4 c3xb4 Bb8d6 b4b5 Kg8f8 b5b6 Bd6b4 b6b7
     31            18:36:54          14.178.750,500           216.685           +3.30  Bc2e4 b5b4 c3xb4 Bb8d6 b4b5 Kg8f8 b5b6
     32            27:29:33          21.123.466,966           218.569           +3.30  Bc2e4 Bb8a7+ Kg1f1
     33+         49:23:23          38.835.699,989           223.675           +3.50  Bc2e4
     33            49:38:04          39.024.361,821           223.652           +3.50  Bc2e4 Bb8a7+ Kg1f1 Kg8f7 Kf1e2
   10/27/2008 2:43:43 AM, Time for this analysis: 02:00:03, Rated time: 50:00:00

0 of 1 matching moves
10/27/2008 2:43:43 AM, Total time: 2:00:18 AM
Rated time: 50:00:00 = 180000 Seconds

[Uri Blass]

Dann,the position with only bishops and pawns on the board that rybka analyzed for 48 hours is not from the game

What are the moves that leaded to the position?

[Uri Blass]

Here is some analysis I did against game 2 positions, some time ago...

[d]1b4k1/6p1/5p1p/Pp1PpP2/2p5/2P4P/2B3P1/6K1 w - - acd 33; acn 39024361821; acs 10720804; bm Be4; ce 350; id "NM(pv)-1j"; pv Be4 Ba7+ Kf1 Kf7 Ke2;

[d]r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - acd 32/87; acn 27740650800; acs 1537202; bm Be4; ce 82; pv Be4 Rcb8 Ra6 Qd8 Kh2 Kh7 g4 Rxa6 Rxa6 Rb7 Kg2 Kg8 Qa2 Kf7 Kg3 Ke7 Qa3 Qc7 Qa1 Kf7 Qg1 Qd8 Kg2 Kg8 Qf2 Kh8 Qa2 Kh7 Kf3 Bb8 Ra8 Qc7 d6 Qd7 Bxb7 Qxb7+ Kg3 Bxd6;

[d]r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - acd 30; acn 4505215255; acs 1123243; bm Be4; ce 60; pv Be4 Qd8 Kh2 Rxa2 Qxa2 Rb8 g3 Kf8 Kg2 Bc7 Qa7 Bb6 Qa6 Bc7 Qc6 Rb6 Qa8;

[d]r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - acd 26/77; acn 6379248338; acs 388854; bm axb5; ce 80; pv axb5 axb5 Be4 Qd8 Kh2 Rcb8 Ra6 Rxa6 Rxa6 Rb7 Qa2 Kh7 Qe2 Kg8 Qf2 Kh8 Qe3 Kg8 g4 Kh7 Qf2 Kh8 Qa2 Rb8 Ra7 Qb6 Kh1 Rd8;

I really don't have the experience, or the software, to decipher this. Is this convincing evidence that a decent program would not allow ...Qe3? Can it be rephrased in chessic terms? My chess is a lot better than my code reading.

It is more likely that simply programs don't understand what is going on. And whether they play Kh1 or something else is based on random luck rather than understanding here.

This draw is _deep_. Several of us spent all night using every computer we had to convince everyone it was really drawn. Until then, even the best player on the planet thought it was lost.

I disagree.

1)The choice of Kh1 is not based on random luck but based on positional evaluation.

If all good programs choose Kh1 after a long search then it is not logical that all good programs have the same luck.

2)I believe that there is no forced draw after the game move Kf1 or after Qe3 that kasparov did not play that team of human and chess programs can see.

I agree to play it against you with white in correspondence game.
I am not sure if I am going to win but I believe that chess programs are going to see an advantage for white for more than 60 plies after the Qe3 line.

[Dann Corbit]

Dann,the position with only bishops and pawns on the board that rybka analyzed for 48 hours is not from the game

What are the moves that leaded to the position?

r1r1q1k1/6p1/1Q1b1p1p/1p1PpP2/1Pp5/2P4P/R1B3P1/R5K1 b - - acd 26; acn -578768639; acs 3600; ce -46; id "NM(pv)-1a"; pv Rxa2 Rxa2 Ra8 Ra5 Rxa5 bxa5 Qb8 Qxb8+ Bxb8 Kf2 Kf7 Be4 Ke7 Kg3 Kd6 Kh4 Kc5 Kh5 Bc7 a6 Kb6 Kg6 Kxa6 Kxg7 Ba5 d6 Bxc3 Kxf6;
2r1q1k1/6p1/1Q1b1p1p/1p1PpP2/1Pp5/2P4P/r1B3P1/R5K1 w - - acd 26; acn -496324613; acs 3601; ce 46; id "NM(pv)-1b"; pv Rxa2 Ra8 Ra5 Rxa5 bxa5 Qb8 Qxb8+ Bxb8 Kf2 Kf7 Be4 Ke7 Kg3 Kd6;
2r1q1k1/6p1/1Q1b1p1p/1p1PpP2/1Pp5/2P4P/R1B3P1/6K1 b - - acd 25; acn -788378947; acs 3600; ce -46; id "NM(pv)-1c"; pv Ra8 Ra5 Rxa5 bxa5 Qb8 Qxb8+ Bxb8 Kf2 Kf7 Kf3 Ke7 Be4 Kd6 Kg4 Kc5 Kh5 Bc7 a6 Kb6 Kg6 Kxa6 Kxg7 Ba5 d6 Bxc3 Kxf6;
r3q1k1/6p1/1Q1b1p1p/1p1PpP2/1Pp5/2P4P/R1B3P1/6K1 w - - acd 25; acn -828817343; acs 3600; ce 50; id "NM(pv)-1d"; pv Ra5 Rxa5 bxa5 Qb8 Qxb8+ Bxb8 Kf2 Kf7 Kf3 Ke7 a6 Kd6 Be4 Kc5 Kg4 Kb6 Kh5 Kxa6 Kg6 Bc7 Kxg7 Ba5 Kxf6 Bxc3 Ke6;
r3q1k1/6p1/1Q1b1p1p/Rp1PpP2/1Pp5/2P4P/2B3P1/6K1 b - - acd 24; acn -462695936; acs 3601; ce -50; id "NM(pv)-1e"; pv Rxa5 bxa5 Qb8 Qxb8+ Bxb8 Kf2 Kf7 Kf3 Ke7 a6 Kd6 Be4 Kc5 Kg4 Kb6 Kh5 Kxa6;
4q1k1/6p1/1Q1b1p1p/rp1PpP2/1Pp5/2P4P/2B3P1/6K1 w - - acd 24; acn -742892860; acs 3600; ce 71; id "NM(pv)-1f"; pv bxa5 Qb8 Qxb8+ Bxb8 Be4 Kf7 Kf2 Ke7 Kg3 Kd6 Kg4 Kc5 Kh5 Bc7 a6 Kb6 Kg6 Kxa6 Kxg7 Ba5 Kxh6 Bxc3 Kg7 b4 Kxf6;
4q1k1/6p1/1Q1b1p1p/Pp1PpP2/2p5/2P4P/2B3P1/6K1 b - - acd 24; acn 92876548; acs 3600; ce -91; id "NM(pv)-1g"; pv Qb8 Qxb8+ Bxb8 Be4 Kf7 Kf2 Ke7 Kg3 Kd6 Kg4 Kc5 Kh5 Bc7 a6 Kb6 Kg6 Kxa6 Kxg7 Ba5 Kxh6 Bxc3 Kg7 b4 g4 b3 Kxf6;
1q4k1/6p1/1Q1b1p1p/Pp1PpP2/2p5/2P4P/2B3P1/6K1 w - - acd 24; acn -704062582; acs 3601; ce 69; id "NM(pv)-1h"; pv Qxb8+ Bxb8 Be4 Kf7 Kf2 Ke7 Kg3 Kd6 Kg4 Kc5 Kh5 b4 a6 bxc3 Kg6 Kd4 Bc2 Kxd5 Kxg7 Kc5 Kxh6 Kb5 Kg6 Kxa6 Kxf6;
1Q4k1/6p1/3b1p1p/Pp1PpP2/2p5/2P4P/2B3P1/6K1 b - - acd 30; acn 1310902651; acs 3601; ce -104; id "NM(pv)-1i"; pv Bxb8 Be4 Kf7 Kf2 Ke7 Kg3 Kd6 Kg4 Kc5 a6 Kb6 Kh5 Kxa6 Kg6 Bc7 Kxg7 Bd8 Kxh6 Ka5 d6 Ka4 Kg7 Kb3 Bc6 Kxc3 Bxb5 Kd4 h4 e4;
1b4k1/6p1/5p1p/Pp1PpP2/2p5/2P4P/2B3P1/6K1 w - - acd 28; acn 1468123504; acs 3600; ce 104; id "NM(pv)-1j"; pv Be4 Kf7 Kf2 Ke7 Kg3 Kd6 Kg4 Kc5 a6 Kb6 Kh5 Kxa6 Kg6 Bc7 Kxg7 Bd8;

[bob]

Here is some analysis I did against game 2 positions, some time ago...

[d]1b4k1/6p1/5p1p/Pp1PpP2/2p5/2P4P/2B3P1/6K1 w - - acd 33; acn 39024361821; acs 10720804; bm Be4; ce 350; id "NM(pv)-1j"; pv Be4 Ba7+ Kf1 Kf7 Ke2;

[d]r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - acd 32/87; acn 27740650800; acs 1537202; bm Be4; ce 82; pv Be4 Rcb8 Ra6 Qd8 Kh2 Kh7 g4 Rxa6 Rxa6 Rb7 Kg2 Kg8 Qa2 Kf7 Kg3 Ke7 Qa3 Qc7 Qa1 Kf7 Qg1 Qd8 Kg2 Kg8 Qf2 Kh8 Qa2 Kh7 Kf3 Bb8 Ra8 Qc7 d6 Qd7 Bxb7 Qxb7+ Kg3 Bxd6;

[d]r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - acd 30; acn 4505215255; acs 1123243; bm Be4; ce 60; pv Be4 Qd8 Kh2 Rxa2 Qxa2 Rb8 g3 Kf8 Kg2 Bc7 Qa7 Bb6 Qa6 Bc7 Qc6 Rb6 Qa8;

[d]r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - acd 26/77; acn 6379248338; acs 388854; bm axb5; ce 80; pv axb5 axb5 Be4 Qd8 Kh2 Rcb8 Ra6 Rxa6 Rxa6 Rb7 Qa2 Kh7 Qe2 Kg8 Qf2 Kh8 Qe3 Kg8 g4 Kh7 Qf2 Kh8 Qa2 Rb8 Ra7 Qb6 Kh1 Rd8;

I really don't have the experience, or the software, to decipher this. Is this convincing evidence that a decent program would not allow ...Qe3? Can it be rephrased in chessic terms? My chess is a lot better than my code reading.

It is more likely that simply programs don't understand what is going on. And whether they play Kh1 or something else is based on random luck rather than understanding here.

This draw is _deep_. Several of us spent all night using every computer we had to convince everyone it was really drawn. Until then, even the best player on the planet thought it was lost.

I disagree.

1)The choice of Kh1 is not based on random luck but based on positional evaluation.

If all good programs choose Kh1 after a long search then it is not logical that all good programs have the same luck.

2)I believe that there is no forced draw after the game move Kf1 or after Qe3 that kasparov did not play that team of human and chess programs can see.

I agree to play it against you with white in correspondence game.
I am not sure if I am going to win but I believe that chess programs are going to see an advantage for white for more than 60 plies after the Qe3 line.

IIRC, at least two programs recognized a draw on Qe3. Granted that we ran them for almost 24 hours to see. I might be able to dig up the Crafty analysis where it found a score of 0 after an impossibly long search. I think Junior may have done the same, but that was so long ago I am not certain... The only issue is "is it a valid draw" because if the program things the wrong side is better, it might favor a draw when in reality it would be avoided by humans. But this was a move in 3 minutes or so, and moving the king move is almost certainly based on pure positional stuff for the king, since we have yet to see a program that thinks white is anything but winning there, regardless of the move played.

[Uri Blass]

I see that you analyzed the position after

37.Qb6 Rxa2 38.Rxa2 Ra8 39.Ra5 Rxa5 40.bxa5 Qb8 41.Qxb8+ Bxb8
and it seems that white is winning but deep blue did not play 37.Qb6 and played 37.Be4 during the game.

[d]r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - 0 37

The discussion here was not about this position but about the final position of the game.

[Uri Blass]

I believe that the program that recognized a draw after Qe3 are wrong and they simply evaluated equal material position as equal when it is practically not equal and white has some positional advantage after 45...Qe3 46.Qxd6 Re8 47.Qc7+ Re7 48.Qc6

[dkl]

I am digging this thread out from the grave, since I want to point out some peculiarity that I've stumbled upon when writing my free book about Neural Networks for Chess ( https://github.com/asdfjkl/neural_network_chess )

As we know, Kasparov was a sore loser and accused Depp Blue of cheating, namely that 7.Be4 (not directly trying to grab a pawn) was a move with human intervention, as no computer could have come up with such a move that ignores material gain. [From todays perspective, a human helping a computer is actually a very funny thought.]

He wrote:

“We saw something that went beyond our wildest expectations of how well a
computer would be able to foresee the long term positional consequences of its
decisions. The machine refused to move to a position that had a decisive short-
term advantage – showing a very human sense of danger [...] Even today, weeks
later, no other chess-playing program in the world has been able to correctly
evaluate the consequences of Deep Blue’s position. (Time, vol. 149, no. 21, 1997.)

In his memoir, Kasparov (Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins) writes:

"On what mattered most, on what really destroyed my composure, I was wrong and owe the Deep Blue team an apology. The moves in game two that left me with a lost position and crushed morale were unique only for the time. Within five years, commercial engines running on standard Intel servers could reproduce all of Deep Blue's best moves, even improving on some of the humanlike moves that so impressed me and everyone else at the time. The engine on my laptop today slightly favors the shockingly humanlike move 37.Be4 from game two in less than ten seconds,"

It's very very late, but then again, it's an apology - which shows some stature.

I was wandering though it that is really true, namely if engines at the time could not find the move, or if Kasparov's impression of computer chess was just tainted due to his relations with Chessbase and Fritz. Fritz was always a notorious pawn grabber.

So I let Rebel 8.0 run for an hour in DosBox, and Rebel came up with Be4 (screenshot in Chapter 3.4 of my book). I confirmed with Ed Schröder that Rebel 8.0 was released in 1996, prior to the (second) match with Deep Blue.

I have not thoroughly investigated on what "1 hour in DosBox with default cycle count" translates to 1996 hardware (1 hour? 10 hours? 100 hours?), but it shows that even a standard PC software program could have come up with the move, let alone a supercomputer like DeepBlue.

I've also checked with older versions of Fritz (5.32), and it was _not_ able to find Be4.

[Jouni]

With SF You can analyse all DB games in couple of minutes thoroughly. Game 2.

Analysis by Stockfish 14.1:

1. +- (2.28): 37.Be4 Rcb8 38.Ra6 Qd8 39.Kh2 Kf8 40.Ra7 Rxa7 41.Qxa7 Rb6 42.Qa2 Qc8 43.Kg3 Bb8 44.Kf3 Kg8 45.Qa5 Qd8 46.h4 Kh7 47.Ke2 Kg8 48.Ke3 Qd6 49.Kf3 Qd8 50.Kg4 Qc7 51.Kg3 Qb7 52.Qa2 Qc8 53.Kh3 Kh7 54.g3 Qb7 55.Qa8 Qd7 56.g4
2. +/= (0.64): 37.Qb6 Rxa2 38.Rxa2 Ra8 39.Ra5 Qf8 40.Be4 Rxa5 41.Qxa5 Qe8 42.g4 h5 43.Kf1 hxg4 44.hxg4 Qb8 45.Qa6 Kf7 46.Qc6 Ke7 47.Ke2 Qe8 48.Qb7+ Qd7 49.Qa6 Qe8 50.Qa7+ Qd7 51.Qg1 Kf8 52.Kd2 Qb7 53.Kc1 Qe7 54.Qb6 Qd7 55.Kb1 Kf7 56.Qe3 Qe7 57.Kb2 Ke8 58.Qb6 Qd7 59.Qg1 Qe7 60.g5 fxg5 61.Qb6 Qd7
3. = (0.00): 37.Rxa8 Rxa8 38.Rxa8 Qxa8 39.Qe1 h5 40.Kf1 Qa3 41.h4 e4 42.Bxe4 Kf8 43.Qd2 Qa7 44.Bf3 Qa1+ 45.Qe1 Qxe1+ 46.Kxe1 Ke7 47.Bxh5 Kd7 48.Kd1 Bg3 49.Kc2 Bxh4 50.Bf7 Bg3 51.Be6+ Kd8 52.Bf7
4. = (0.00): 37.Qg3 Rxa2 38.Rxa2 Ra8 39.Rxa8 Qxa8 40.Qe1 h5 41.Kf1 Qa3 42.h4 e4 43.Bxe4 Kf8 44.Qd2 Qa7 45.Bf3 Qa1+ 46.Qe1 Qxe1+ 47.Kxe1 Ke7 48.Bxh5 Kd7 49.Kd1 Bg3 50.Kc2 Bxh4 51.Bf7 Bg3 52.Be6+ Kd8 53.Bf7 Bd6 54.Be6 Be5 55.Bg8 Bf4 56.Bf7

In game 1. Deep Blue errors:

22. -g4?
34. -Qxf1?
36. -Kf8?

[towforce]

In his memoir, Kasparov (Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins) writes:

"On what mattered most, on what really destroyed my composure, I was wrong and owe the Deep Blue team an apology. The moves in game two that left me with a lost position and crushed morale were unique only for the time. Within five years, commercial engines running on standard Intel servers could reproduce all of Deep Blue's best moves, even improving on some of the humanlike moves that so impressed me and everyone else at the time. The engine on my laptop today slightly favors the shockingly humanlike move 37.Be4 from game two in less than ten seconds,"

It's very very late, but then again, it's an apology - which shows some stature.

The apology, no matter how belated, is gracious, and Kasparov deserves a bit of credit for offering it.