Stockfish 1 with 12 threads and 2 hours vs Stockfish 17.1 with 12 threads and 1 minute, who wins?

[Leto]

An old discussion people have had is if you give a weaker but solid engine enough time would it be able to match the strength of a much stronger engine? So I thought it'd be interesting to do a little experiment and do a game between Stockfish 1 with 12 threads and 2 hours vs Stockfish 17.1 with 12 threads and just 1 minute. Before I give you the game let me quickly give some info on the history of these two engines:

Stockfish 1 broke into the scene in 2008 and in 2011 they released the final and strongest version of Stockfish 1x series, Stockfish 1.9.1. CCRL rated it at 3148 in their 40/15 list when running on 4 threads. I know that CCRL ratings aren't the same as FIDE ratings but I'm guessing that if it runs on 12 threads it should be playing around the 2900 FIDE range if given 2 hours for a game.

Stockfish 17.1 was released earlier this year and on 4 threads it is rated 3644 in CCRL's 40/15 list. So that's almost 500 elo stronger than Stockfish 1.9.1.

So if there’s a 500-Elo gap between them, could Stockfish 1.9.1 beat Stockfish 17.1 by having 120 times more time to think? So the old engine gets 2 hours versus the new engine getting only 1 minute for the entire game. I put it to a test and here's what I got:

[Event "\[03\] Engine Match (01/11/2025)"]
[Site "Local computer"]
[Date "2025.11.01"]
[Round "1"]
[White "Stockfish 1.9.1 JA 64bit 12CPU"]
[Black "Stockfish 17.1 12CPU"]
[Result "0-1"]
[ECO "D38"]
[PlyCount "160"]
[EventDate "2025.??.??"]

1. d4 Nf6 2. Nf3 e6 3. c4 d5 4. Nc3 Bb4 5. Qa4+ Nc6 6. e3 O-O 7. Bd2 dxc4 8. Bxc4 Bd6 9. Nb5 Bd7 10. Nxd6 cxd6 11. Qa3 Rc8 12. Rc1 d5 13. Bd3 Ne4 14. O-O Nxd2 15. Nxd2 Qe7 16. Qxe7 Nxe7 17. f3 Rxc1 18. Rxc1 Rc8 19. Rxc8+ Nxc8 20. Kf2 Nb6 21. b3 Kf8 22. e4 dxe4 23. fxe4 h6 24. Ke3 f6 25. a4 Ke7 26. Nb1 e5 27. Nd2 Nc8 28. d5 Kd6 29. Nf3 Kc5 30. Bc2 b5 31. axb5 Nd6 32. b6 axb6 33. h3 Kb4 34. Nd2 Kc3 35. Bd3 b5 36. Bb1 Nb7 37. Bd3 Nd6 38. Bb1 h5 39. Bd3 h4 40. Be2 f5 41. Bd1 f4+ 42. Ke2 b4 43. Ke1 Bb5 44. Nf3 Bd3 45. Nxe5 Bxe4 46. Bf3 Bc2 47. Bd1 Kxb3 48. Ng6 Kb2 49. Nxf4 b3 50. Ke2 Kc3 51. Bxc2 Kxc2 52. Nd3 Nc4 53. Nb4+ Kc3 54. Nd3 Kd4 55. d6 Nxd6 56. Nb2 Nc4 57. Na4 Ne5 58. Kd2 g5 59. Kd1 Nc4 60. Ke2 b2 61. Nxb2 Nxb2 62. Kf3 Nc4 63. Kf2 Kd3 64. g3 Nd2 65. g4 Nc4 66. Ke1 Ke3 67. Kd1 Ne5 68. Ke1 Nd3+ 69. Kd1 Nf4 70. Ke1 Nxh3 71. Kf1 Nf2 72. Kg2 Nxg4 73. Kh3 Nf2+ 74. Kg2 g4 75. Kg1 Kf3 76. Kf1 Nd3 77. Kg1 g3 78. Kf1 h3 79. Kg1 h2+ 80. Kh1 Nf2# 0-1

Note that neither side had a book and when the game ended Stockfish 1.9.1 still had just under 6 minutes left on its clock and Stockfish 17.1 just had milliseconds remaining. With just 1 minute for the entire game it's an amazing achievement by Stockfish 17.1.

I did a game review here at chess.com using max settings and it gave Stockfish 1.9.1 an accuracy score of 91.5 and a game rating of 2450 and Stockfish 17.1 got 98.2 and a game rating of 2500 (it seems to max out at 2500).

Here's chess.com's move rating counts for this game, for white you'll see the 'w' letter next to the count, for black you'll see 'b' next to the count:

Brilliant
w0
0b


Great

w1

1b
Best
w42
58b

Excellent
w18
11b

Good
w4
2b

Book
w8
8b


Inaccuracy

w3
0b

Mistake
w2
0b

Miss
w0
0b

Blunder
w0
0b


According to the game review white's inaccuracies were 35.Bd3 (-2.61), 43.Ke1 (-4.11), and 64.g3 (-M20), and white's mistakes were 28.d5 (-0.97) and 33.h3 (-1.66). Let's take a closer look at these errors:

35.Bd3 (-2.61). Better is 35.Bd1 but maybe it's already lost as the score is down to -2.91 according to cloud Stockfish 16 at depth 49. Out of curiosity I looked at what cloud SF16 thinks of the played move 35.Bd1 and at depth 44 it's at -3.22.

43.Ke1 (-4.11). No this might be the best move, at depth 41 cloud SF16 says this is best with a score of -4.87.

64.g3 (-M20). Well 64.Kf3 is cloud SF16's top move at depth 82 with a score of -M19. I had it look at 64.g3 and it says -M18.

28.d5 (-0.97). Cloud SF16 says 28.a5 is best and 0.0 at depth 58.

33.h3 (-1.66). Cloud SF16 says 33.Nh4 is best and -0.92 at depth 50

So in conclusion I don't think giving an old engine enough time would ever make it as strong as today's top engines. Today's engines have better understanding of all phases of the game. Giving an old engine more time will increase its strength but only to a certain limit, possibly around 400 elo maximum increase.

[Ajedrecista]

Hello:

Thank you for the game. I enclose it between PGN tags to get a better view of the game:

[pgn][Event "\[03\] Engine Match (01/11/2025)"]
[Site "Local computer"]
[Date "2025.11.01"]
[Round "1"]
[White "Stockfish 1.9.1 JA 64bit 12CPU"]
[Black "Stockfish 17.1 12CPU"]
[Result "0-1"]
[ECO "D38"]
[PlyCount "160"]
[EventDate "2025.??.??"]

1. d4 Nf6 2. Nf3 e6 3. c4 d5 4. Nc3 Bb4 5. Qa4+ Nc6 6. e3 O-O 7. Bd2 dxc4 8. Bxc4 Bd6 9. Nb5 Bd7 10. Nxd6 cxd6 11. Qa3 Rc8 12. Rc1 d5 13. Bd3 Ne4 14. O-O Nxd2 15. Nxd2 Qe7 16. Qxe7 Nxe7 17. f3 Rxc1 18. Rxc1 Rc8 19. Rxc8+ Nxc8 20. Kf2 Nb6 21. b3 Kf8 22. e4 dxe4 23. fxe4 h6 24. Ke3 f6 25. a4 Ke7 26. Nb1 e5 27. Nd2 Nc8 28. d5 Kd6 29. Nf3 Kc5 30. Bc2 b5 31. axb5 Nd6 32. b6 axb6 33. h3 Kb4 34. Nd2 Kc3 35. Bd3 b5 36. Bb1 Nb7 37. Bd3 Nd6 38. Bb1 h5 39. Bd3 h4 40. Be2 f5 41. Bd1 f4+ 42. Ke2 b4 43. Ke1 Bb5 44. Nf3 Bd3 45. Nxe5 Bxe4 46. Bf3 Bc2 47. Bd1 Kxb3 48. Ng6 Kb2 49. Nxf4 b3 50. Ke2 Kc3 51. Bxc2 Kxc2 52. Nd3 Nc4 53. Nb4+ Kc3 54. Nd3 Kd4 55. d6 Nxd6 56. Nb2 Nc4 57. Na4 Ne5 58. Kd2 g5 59. Kd1 Nc4 60. Ke2 b2 61. Nxb2 Nxb2 62. Kf3 Nc4 63. Kf2 Kd3 64. g3 Nd2 65. g4 Nc4 66. Ke1 Ke3 67. Kd1 Ne5 68. Ke1 Nd3+ 69. Kd1 Nf4 70. Ke1 Nxh3 71. Kf1 Nf2 72. Kg2 Nxg4 73. Kh3 Nf2+ 74. Kg2 g4 75. Kg1 Kf3 76. Kf1 Nd3 77. Kg1 g3 78. Kf1 h3 79. Kg1 h2+ 80. Kh1 Nf2# 0-1[/pgn]

A quick analysis by SF 17 retrieved a bunch of moves with 0 or almost 0 evals, but not 28.- d5?, which seems to be the start of the nightmare, indeed. 28.- ..., Kd6 is the only move with real winning chances for black.

The current ratings of Stockfish 17.1 64-bit 4CPU and Stockfish 1.9.1 64-bit 4CPU at CCRL 40/15 complete list are:

Code: Select all

                                       Rating                     Average 
Name	                        Elo       +       -      Score    Opponent   Draws    Games	
Stockfish 17.1  64-bit 4CPU     3644     +11     −11     57.1%     −43.7     85.1%     2012
Stockfish 1.9.1 64-bit 4CPU     3148     +17     −17     58.0%     −58.7     47.0%      809

For a difference of 496 Elo, error bars aside, as you already stated above. I know, 4 CPU each instead of 12.

OTOH, an old study used a formula to determine thinking times starting from Elo differences:

Code: Select all

log2[ t(A) / t(B) ] = [ Elo(B) - Elo(A) ] / 120

The constant 120 makes the difference, meaning each doubling in TC worths 120 Elo. We could talk about diminishing returns and complicate the subject further.

Anyway, with a 496 Elo difference: t(SF 1.9.1) / t(SF 17.1) = 2^(496/120) ~ 17.55 to get an even match, which did not prove enough for SF 1.9.1 in your isolated game [t(SF 1.9.1) / t(SF 17.1) = 120 > 17.55]. Just changing 120 by 62 (more in line with other estimates of 50 to 70 Elo per doubling of cores, might not per doubling of time): t(SF 1.9.1) / t(SF 17.1) = 2^(496/62) = 256 < 120 of the game. 62 was not chosen at random, I wanted to get an integer exponent.

496 / log2(120) ~ 71.81 Elo per doubling in time. As you see, tiny changes in the constant make large changes in time ratios. How about a rematch with 2 hours versus 30 seconds, for a time ratio of 240? It could be probably enough for SF 17.1 to repeat the feat again. A try to fool SF 17.1 would be lowering its available time too much, for example 1 second for the whole game and 4 minutes for SF 1.9.1 to kkep the 240:1 time ratio, with the possibility of play more games.

Regards from Spain.

Ajedrecista.

[Uri Blass]

An old discussion people have had is if you give a weaker but solid engine enough time would it be able to match the strength of a much stronger engine? So I thought it'd be interesting to do a little experiment and do a game between Stockfish 1 with 12 threads and 2 hours vs Stockfish 17.1 with 12 threads and just 1 minute. Before I give you the game let me quickly give some info on the history of these two engines:

Stockfish 1 broke into the scene in 2008 and in 2011 they released the final and strongest version of Stockfish 1x series, Stockfish 1.9.1. CCRL rated it at 3148 in their 40/15 list when running on 4 threads. I know that CCRL ratings aren't the same as FIDE ratings but I'm guessing that if it runs on 12 threads it should be playing around the 2900 FIDE range if given 2 hours for a game.

Stockfish 17.1 was released earlier this year and on 4 threads it is rated 3644 in CCRL's 40/15 list. So that's almost 500 elo stronger than Stockfish 1.9.1.

So if there’s a 500-Elo gap between them, could Stockfish 1.9.1 beat Stockfish 17.1 by having 120 times more time to think? So the old engine gets 2 hours versus the new engine getting only 1 minute for the entire game. I put it to a test and here's what I got:

[Event "\[03\] Engine Match (01/11/2025)"]
[Site "Local computer"]
[Date "2025.11.01"]
[Round "1"]
[White "Stockfish 1.9.1 JA 64bit 12CPU"]
[Black "Stockfish 17.1 12CPU"]
[Result "0-1"]
[ECO "D38"]
[PlyCount "160"]
[EventDate "2025.??.??"]

1. d4 Nf6 2. Nf3 e6 3. c4 d5 4. Nc3 Bb4 5. Qa4+ Nc6 6. e3 O-O 7. Bd2 dxc4 8. Bxc4 Bd6 9. Nb5 Bd7 10. Nxd6 cxd6 11. Qa3 Rc8 12. Rc1 d5 13. Bd3 Ne4 14. O-O Nxd2 15. Nxd2 Qe7 16. Qxe7 Nxe7 17. f3 Rxc1 18. Rxc1 Rc8 19. Rxc8+ Nxc8 20. Kf2 Nb6 21. b3 Kf8 22. e4 dxe4 23. fxe4 h6 24. Ke3 f6 25. a4 Ke7 26. Nb1 e5 27. Nd2 Nc8 28. d5 Kd6 29. Nf3 Kc5 30. Bc2 b5 31. axb5 Nd6 32. b6 axb6 33. h3 Kb4 34. Nd2 Kc3 35. Bd3 b5 36. Bb1 Nb7 37. Bd3 Nd6 38. Bb1 h5 39. Bd3 h4 40. Be2 f5 41. Bd1 f4+ 42. Ke2 b4 43. Ke1 Bb5 44. Nf3 Bd3 45. Nxe5 Bxe4 46. Bf3 Bc2 47. Bd1 Kxb3 48. Ng6 Kb2 49. Nxf4 b3 50. Ke2 Kc3 51. Bxc2 Kxc2 52. Nd3 Nc4 53. Nb4+ Kc3 54. Nd3 Kd4 55. d6 Nxd6 56. Nb2 Nc4 57. Na4 Ne5 58. Kd2 g5 59. Kd1 Nc4 60. Ke2 b2 61. Nxb2 Nxb2 62. Kf3 Nc4 63. Kf2 Kd3 64. g3 Nd2 65. g4 Nc4 66. Ke1 Ke3 67. Kd1 Ne5 68. Ke1 Nd3+ 69. Kd1 Nf4 70. Ke1 Nxh3 71. Kf1 Nf2 72. Kg2 Nxg4 73. Kh3 Nf2+ 74. Kg2 g4 75. Kg1 Kf3 76. Kf1 Nd3 77. Kg1 g3 78. Kf1 h3 79. Kg1 h2+ 80. Kh1 Nf2# 0-1

Note that neither side had a book and when the game ended Stockfish 1.9.1 still had just under 6 minutes left on its clock and Stockfish 17.1 just had milliseconds remaining. With just 1 minute for the entire game it's an amazing achievement by Stockfish 17.1.

I did a game review here at chess.com using max settings and it gave Stockfish 1.9.1 an accuracy score of 91.5 and a game rating of 2450 and Stockfish 17.1 got 98.2 and a game rating of 2500 (it seems to max out at 2500).

Here's chess.com's move rating counts for this game, for white you'll see the 'w' letter next to the count, for black you'll see 'b' next to the count:

Brilliant
w0
0b


Great

w1

1b
Best
w42
58b

Excellent
w18
11b

Good
w4
2b

Book
w8
8b


Inaccuracy

w3
0b

Mistake
w2
0b

Miss
w0
0b

Blunder
w0
0b


According to the game review white's inaccuracies were 35.Bd3 (-2.61), 43.Ke1 (-4.11), and 64.g3 (-M20), and white's mistakes were 28.d5 (-0.97) and 33.h3 (-1.66). Let's take a closer look at these errors:

35.Bd3 (-2.61). Better is 35.Bd1 but maybe it's already lost as the score is down to -2.91 according to cloud Stockfish 16 at depth 49. Out of curiosity I looked at what cloud SF16 thinks of the played move 35.Bd1 and at depth 44 it's at -3.22.

43.Ke1 (-4.11). No this might be the best move, at depth 41 cloud SF16 says this is best with a score of -4.87.

64.g3 (-M20). Well 64.Kf3 is cloud SF16's top move at depth 82 with a score of -M19. I had it look at 64.g3 and it says -M18.

28.d5 (-0.97). Cloud SF16 says 28.a5 is best and 0.0 at depth 58.

33.h3 (-1.66). Cloud SF16 says 33.Nh4 is best and -0.92 at depth 50

So in conclusion I don't think giving an old engine enough time would ever make it as strong as today's top engines. Today's engines have better understanding of all phases of the game. Giving an old engine more time will increase its strength but only to a certain limit, possibly around 400 elo maximum increase.

The question is if stockfish can avoid 28.d5 if you give it enough time per move.
Did you try to analyze with stockfish for 24 hours when it still suggested 28.d5?

[Uri Blass]

Note that I do not see where to download stockfish1.9.1 to test it in the relevant position

[syzygy]

Note that I do not see where to download stockfish1.9.1 to test it in the relevant position

If you know how to compile it yourself:
https://github.com/official-stockfish/S ... e/sf_1.9.1

[acase]

Note that I do not see where to download stockfish1.9.1 to test it in the relevant position

Try here: https://stockfishchess.org/download/

Scroll down to where it says "Old (archived) releases of Stockfish"

It's at the bottom of the page.

[Leto]

Hello:

Thank you for the game. I enclose it between PGN tags to get a better view of the game:

[pgn][Event "\[03\] Engine Match (01/11/2025)"]
[Site "Local computer"]
[Date "2025.11.01"]
[Round "1"]
[White "Stockfish 1.9.1 JA 64bit 12CPU"]
[Black "Stockfish 17.1 12CPU"]
[Result "0-1"]
[ECO "D38"]
[PlyCount "160"]
[EventDate "2025.??.??"]

1. d4 Nf6 2. Nf3 e6 3. c4 d5 4. Nc3 Bb4 5. Qa4+ Nc6 6. e3 O-O 7. Bd2 dxc4 8. Bxc4 Bd6 9. Nb5 Bd7 10. Nxd6 cxd6 11. Qa3 Rc8 12. Rc1 d5 13. Bd3 Ne4 14. O-O Nxd2 15. Nxd2 Qe7 16. Qxe7 Nxe7 17. f3 Rxc1 18. Rxc1 Rc8 19. Rxc8+ Nxc8 20. Kf2 Nb6 21. b3 Kf8 22. e4 dxe4 23. fxe4 h6 24. Ke3 f6 25. a4 Ke7 26. Nb1 e5 27. Nd2 Nc8 28. d5 Kd6 29. Nf3 Kc5 30. Bc2 b5 31. axb5 Nd6 32. b6 axb6 33. h3 Kb4 34. Nd2 Kc3 35. Bd3 b5 36. Bb1 Nb7 37. Bd3 Nd6 38. Bb1 h5 39. Bd3 h4 40. Be2 f5 41. Bd1 f4+ 42. Ke2 b4 43. Ke1 Bb5 44. Nf3 Bd3 45. Nxe5 Bxe4 46. Bf3 Bc2 47. Bd1 Kxb3 48. Ng6 Kb2 49. Nxf4 b3 50. Ke2 Kc3 51. Bxc2 Kxc2 52. Nd3 Nc4 53. Nb4+ Kc3 54. Nd3 Kd4 55. d6 Nxd6 56. Nb2 Nc4 57. Na4 Ne5 58. Kd2 g5 59. Kd1 Nc4 60. Ke2 b2 61. Nxb2 Nxb2 62. Kf3 Nc4 63. Kf2 Kd3 64. g3 Nd2 65. g4 Nc4 66. Ke1 Ke3 67. Kd1 Ne5 68. Ke1 Nd3+ 69. Kd1 Nf4 70. Ke1 Nxh3 71. Kf1 Nf2 72. Kg2 Nxg4 73. Kh3 Nf2+ 74. Kg2 g4 75. Kg1 Kf3 76. Kf1 Nd3 77. Kg1 g3 78. Kf1 h3 79. Kg1 h2+ 80. Kh1 Nf2# 0-1[/pgn]

A quick analysis by SF 17 retrieved a bunch of moves with 0 or almost 0 evals, but not 28.- d5?, which seems to be the start of the nightmare, indeed. 28.- ..., Kd6 is the only move with real winning chances for black.

The current ratings of Stockfish 17.1 64-bit 4CPU and Stockfish 1.9.1 64-bit 4CPU at CCRL 40/15 complete list are:

Code: Select all

                                       Rating                     Average 
Name	                        Elo       +       -      Score    Opponent   Draws    Games	
Stockfish 17.1  64-bit 4CPU     3644     +11     −11     57.1%     −43.7     85.1%     2012
Stockfish 1.9.1 64-bit 4CPU     3148     +17     −17     58.0%     −58.7     47.0%      809

For a difference of 496 Elo, error bars aside, as you already stated above. I know, 4 CPU each instead of 12.

OTOH, an old study used a formula to determine thinking times starting from Elo differences:

Code: Select all

log2[ t(A) / t(B) ] = [ Elo(B) - Elo(A) ] / 120

The constant 120 makes the difference, meaning each doubling in TC worths 120 Elo. We could talk about diminishing returns and complicate the subject further.

Anyway, with a 496 Elo difference: t(SF 1.9.1) / t(SF 17.1) = 2^(496/120) ~ 17.55 to get an even match, which did not prove enough for SF 1.9.1 in your isolated game [t(SF 1.9.1) / t(SF 17.1) = 120 > 17.55]. Just changing 120 by 62 (more in line with other estimates of 50 to 70 Elo per doubling of cores, might not per doubling of time): t(SF 1.9.1) / t(SF 17.1) = 2^(496/62) = 256 < 120 of the game. 62 was not chosen at random, I wanted to get an integer exponent.

496 / log2(120) ~ 71.81 Elo per doubling in time. As you see, tiny changes in the constant make large changes in time ratios. How about a rematch with 2 hours versus 30 seconds, for a time ratio of 240? It could be probably enough for SF 17.1 to repeat the feat again. A try to fool SF 17.1 would be lowering its available time too much, for example 1 second for the whole game and 4 minutes for SF 1.9.1 to kkep the 240:1 time ratio, with the possibility of play more games.

Regards from Spain.

Ajedrecista.

Thanks I might do 1 minute vs 4 hours.

[Uri Blass]

Note that I do not see where to download stockfish1.9.1 to test it in the relevant position

Try here: https://stockfishchess.org/download/

Scroll down to where it says "Old (archived) releases of Stockfish"

It's at the bottom of the page.

thanks

[Leto]

An old discussion people have had is if you give a weaker but solid engine enough time would it be able to match the strength of a much stronger engine? So I thought it'd be interesting to do a little experiment and do a game between Stockfish 1 with 12 threads and 2 hours vs Stockfish 17.1 with 12 threads and just 1 minute. Before I give you the game let me quickly give some info on the history of these two engines:

Stockfish 1 broke into the scene in 2008 and in 2011 they released the final and strongest version of Stockfish 1x series, Stockfish 1.9.1. CCRL rated it at 3148 in their 40/15 list when running on 4 threads. I know that CCRL ratings aren't the same as FIDE ratings but I'm guessing that if it runs on 12 threads it should be playing around the 2900 FIDE range if given 2 hours for a game.

Stockfish 17.1 was released earlier this year and on 4 threads it is rated 3644 in CCRL's 40/15 list. So that's almost 500 elo stronger than Stockfish 1.9.1.

So if there’s a 500-Elo gap between them, could Stockfish 1.9.1 beat Stockfish 17.1 by having 120 times more time to think? So the old engine gets 2 hours versus the new engine getting only 1 minute for the entire game. I put it to a test and here's what I got:

[Event "\[03\] Engine Match (01/11/2025)"]
[Site "Local computer"]
[Date "2025.11.01"]
[Round "1"]
[White "Stockfish 1.9.1 JA 64bit 12CPU"]
[Black "Stockfish 17.1 12CPU"]
[Result "0-1"]
[ECO "D38"]
[PlyCount "160"]
[EventDate "2025.??.??"]

1. d4 Nf6 2. Nf3 e6 3. c4 d5 4. Nc3 Bb4 5. Qa4+ Nc6 6. e3 O-O 7. Bd2 dxc4 8. Bxc4 Bd6 9. Nb5 Bd7 10. Nxd6 cxd6 11. Qa3 Rc8 12. Rc1 d5 13. Bd3 Ne4 14. O-O Nxd2 15. Nxd2 Qe7 16. Qxe7 Nxe7 17. f3 Rxc1 18. Rxc1 Rc8 19. Rxc8+ Nxc8 20. Kf2 Nb6 21. b3 Kf8 22. e4 dxe4 23. fxe4 h6 24. Ke3 f6 25. a4 Ke7 26. Nb1 e5 27. Nd2 Nc8 28. d5 Kd6 29. Nf3 Kc5 30. Bc2 b5 31. axb5 Nd6 32. b6 axb6 33. h3 Kb4 34. Nd2 Kc3 35. Bd3 b5 36. Bb1 Nb7 37. Bd3 Nd6 38. Bb1 h5 39. Bd3 h4 40. Be2 f5 41. Bd1 f4+ 42. Ke2 b4 43. Ke1 Bb5 44. Nf3 Bd3 45. Nxe5 Bxe4 46. Bf3 Bc2 47. Bd1 Kxb3 48. Ng6 Kb2 49. Nxf4 b3 50. Ke2 Kc3 51. Bxc2 Kxc2 52. Nd3 Nc4 53. Nb4+ Kc3 54. Nd3 Kd4 55. d6 Nxd6 56. Nb2 Nc4 57. Na4 Ne5 58. Kd2 g5 59. Kd1 Nc4 60. Ke2 b2 61. Nxb2 Nxb2 62. Kf3 Nc4 63. Kf2 Kd3 64. g3 Nd2 65. g4 Nc4 66. Ke1 Ke3 67. Kd1 Ne5 68. Ke1 Nd3+ 69. Kd1 Nf4 70. Ke1 Nxh3 71. Kf1 Nf2 72. Kg2 Nxg4 73. Kh3 Nf2+ 74. Kg2 g4 75. Kg1 Kf3 76. Kf1 Nd3 77. Kg1 g3 78. Kf1 h3 79. Kg1 h2+ 80. Kh1 Nf2# 0-1

Note that neither side had a book and when the game ended Stockfish 1.9.1 still had just under 6 minutes left on its clock and Stockfish 17.1 just had milliseconds remaining. With just 1 minute for the entire game it's an amazing achievement by Stockfish 17.1.

I did a game review here at chess.com using max settings and it gave Stockfish 1.9.1 an accuracy score of 91.5 and a game rating of 2450 and Stockfish 17.1 got 98.2 and a game rating of 2500 (it seems to max out at 2500).

Here's chess.com's move rating counts for this game, for white you'll see the 'w' letter next to the count, for black you'll see 'b' next to the count:

Brilliant
w0
0b


Great

w1

1b
Best
w42
58b

Excellent
w18
11b

Good
w4
2b

Book
w8
8b


Inaccuracy

w3
0b

Mistake
w2
0b

Miss
w0
0b

Blunder
w0
0b


According to the game review white's inaccuracies were 35.Bd3 (-2.61), 43.Ke1 (-4.11), and 64.g3 (-M20), and white's mistakes were 28.d5 (-0.97) and 33.h3 (-1.66). Let's take a closer look at these errors:

35.Bd3 (-2.61). Better is 35.Bd1 but maybe it's already lost as the score is down to -2.91 according to cloud Stockfish 16 at depth 49. Out of curiosity I looked at what cloud SF16 thinks of the played move 35.Bd1 and at depth 44 it's at -3.22.

43.Ke1 (-4.11). No this might be the best move, at depth 41 cloud SF16 says this is best with a score of -4.87.

64.g3 (-M20). Well 64.Kf3 is cloud SF16's top move at depth 82 with a score of -M19. I had it look at 64.g3 and it says -M18.

28.d5 (-0.97). Cloud SF16 says 28.a5 is best and 0.0 at depth 58.

33.h3 (-1.66). Cloud SF16 says 33.Nh4 is best and -0.92 at depth 50

So in conclusion I don't think giving an old engine enough time would ever make it as strong as today's top engines. Today's engines have better understanding of all phases of the game. Giving an old engine more time will increase its strength but only to a certain limit, possibly around 400 elo maximum increase.

The question is if stockfish can avoid 28.d5 if you give it enough time per move.
Did you try to analyze with stockfish for 24 hours when it still suggested 28.d5?

From the notes I have in the actual game Stockfish 1.9.1 took 235 seconds to make 28.d5 and it scored it 0.00 at depth 37. I analyzed the position with SF 1.9.1 with 12 threads and it did manage to find 28.a5 is best with a score of +0.08 after just 208 seconds and depth 33 so I think it was just bad luck that it didn't find it in the actual game.

Here's the analysis it gave me: 28. a5 exd4+ 29. Kxd4 Nd6 30. e5 fxe5+ 31. Kxe5 Nf7+ 32. Kd4 Nd6 33. g3 Ke6
34. Nc4 Nxc4 35. Bxc4+ Kd6 36. Bd5 Bc6 37. Bf7 Bf3 38. Bc4 Bg2 39. b4 Bc6
{(0:03:28) 2671414kN}{[%eval 8,33,Stockfish 1.9.1 JA 64bit 12CPU][%meval 208s]}
*

I also see that it first noticed that after 28.a5 Kd6 is best after 153s. Just bad luck I guess in this case which I think would be fixed if given more time, at least in this case.

28. d5 Kd6 {(0:02:33) 2029943kN}{[%eval 4,33,Stockfish 1.9.1 JA 64bit
12CPU][%meval 153s]} *

Before noticing Kd6 it thought 28.d5 Nd6 was best for black.

[Uri Blass]

This is with 6 threads
[d]2n5/pp1bk1p1/5p1p/4p3/P2PP3/1P1BK3/3N2PP/8 w - - 2 28
Stockfish-191-64-ja:
1 00:00 258 258k +1.33 d4-d5
2 00:00 304 304k +0.48 d4-d5 Nc8-d6
2 00:00 376 376k +0.52 Nd2-f3 Bd7-e6
2 00:00 406 406k +0.64 d4xe5 f6xe5
3 00:00 685 685k +0.84 d4xe5 f6xe5 Nd2-f3
4 00:00 1k 543k +0.84 d4xe5 f6xe5 Nd2-c4 Bd7-e6
5 00:00 1k 725k +0.76 d4xe5 f6xe5 Nd2-c4 Bd7-e6 a4-a5
6- 00:00 2k 618k +0.68 d4xe5 f6xe5 Nd2-c4 Bd7-e6 a4-a5 Nc8-d6
6 00:00 2k 802k +0.64 d4xe5 f6xe5 Nd2-c4 Bd7-e6 Nc4xe5 Be6xb3
7- 00:00 5k 706k +0.48 d4xe5 f6xe5 Nd2-c4 Bd7-e6 Nc4xe5 Be6xb3
7- 00:00 8k 766k +0.32 d4xe5 f6xe5 Nd2-c4 Bd7-e6 Nc4xe5 Be6xb3
7 00:00 10k 677k +0.44 d4-d5 Nc8-d6 h2-h3 a7-a6 Nd2-c4 Nd6xc4+ Bd3xc4 Ke7-d6 Bc4-e2
8 00:00 34k 906k +0.36 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 Be8-g6 Nf3-h4 Bg6-e8 g2-g4
9 00:00 36k 828k +0.36 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 Be8-g6 Nf3-h4 Bg6-e8 g2-g4
10 00:00 52k 864k +0.36 d4-d5 Nc8-d6 h2-h3 a7-a6 g2-g4 Bd7-e8 Nd2-f3 Be8-g6 Nf3-h4 Bg6-h7 Nh4-f5+ Bh7xf5 g4xf5
11- 00:00 68k 916k +0.28 d4-d5 Nc8-d6 h2-h3 a7-a6 g2-g4 Bd7-e8 Nd2-f3 Be8-g6 Nf3-h4 Bg6-h7 Nh4-f5+ Bh7xf5 g4xf5 Ke7-d7
11 00:00 112k 1,074k +0.28 d4-d5 Nc8-d6 Nd2-f3 a7-a6 a4-a5 Bd7-e8 g2-g4 Be8-g6 Nf3-h4 Bg6-h7 Nh4-f5+ Bh7xf5 g4xf5
12 00:00 182k 1,208k +0.24 d4-d5 Nc8-d6 Nd2-f3 b7-b5 a4xb5 Bd7xb5 Bd3xb5 Nd6xb5 g2-g4 Nb5-d6 h2-h3 a7-a6 Nf3-h4 Ke7-d7 Nh4-f5 Nd6xf5+ g4xf5 Kd7-d6
13+ 00:00 246k 1,318k +0.32 d4-d5 Nc8-d6 Nd2-f3 b7-b5 a4xb5 Bd7xb5 Bd3xb5 Nd6xb5 g2-g4 Nb5-d6 h2-h3 a7-a6 Nf3-d2 Ke7-d7 Nd2-c4
13 00:00 248k 1,325k +0.32 d4-d5 Nc8-d6 Nd2-f3 b7-b5 a4xb5 Bd7xb5 Bd3xb5 Nd6xb5 g2-g4 Nb5-d6 h2-h3 a7-a6 Nf3-d2 Ke7-d7 Nd2-c4
14 00:00 315k 1,406k +0.32 d4-d5 Nc8-d6 Nd2-f3 b7-b5 a4xb5 Bd7xb5 Bd3xb5 Nd6xb5 g2-g4 Nb5-d6 h2-h3 Ke7-d7 Nf3-d2 Kd7-c7 Ke3-d3 Kc7-d7 Nd2-c4
15 00:00 391k 1,471k +0.32 d4-d5 Nc8-d6 Nd2-f3 b7-b5 a4xb5 Bd7xb5 Bd3xb5 Nd6xb5 g2-g4 Nb5-d6 h2-h3 Ke7-d7 Nf3-h4 g7-g5 Nh4-f3 a7-a6 Nf3-d2 a6-a5
16 00:00 645k 1,773k +0.28 d4-d5 Nc8-d6 Nd2-f3 b7-b5 a4xb5 Bd7xb5 Bd3xb5 Nd6xb5 g2-g4 Nb5-d6 Ke3-d3 Nd6-b7 Nf3-d2 Nb7-d6 Nd2-c4 Nd6-b7 b3-b4 Nb7-d6 h2-h3
17- 00:02 2,419k 934k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 g2-g4 a7-a6 Nf3-h4 g7-g5 Nh4-f5+ Nd6xf5+ g4xf5 Ke7-d6 h2-h3 Be8-h5 a4-a5 Bh5-e8 Bd3-e2 Be8-d7
17 00:03 3,069k 905k +0.24 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 g7-g5 Nh4-f3 Be8-g6 g2-g3 a7-a6 Nf3-d2 Bg6-h5 h2-h3 Bh5-g6 g3-g4 h6-h5 Nd2-c4 h5xg4 h3xg4
18 00:03 3,223k 867k +0.24 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 g7-g5 Nh4-f3 Be8-g6 g2-g3 a7-a6 Nf3-d2 Bg6-h5 h2-h3 Bh5-g6 Nd2-f3 h6-h5 a4-a5 Ke7-d7 b3-b4 Kd7-e7 Nf3-d2 h5-h4
19 00:04 3,624k 852k +0.24 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 g7-g5 Nh4-f3 Be8-g6 g2-g3 a7-a6 Nf3-d2 f6-f5 e4xf5 Nd6xf5+ Bd3xf5 Bg6xf5 Nd2-e4 Ke7-d7 a4-a5 Bf5-g6 h2-h4 Bg6xe4 Ke3xe4 Kd7-d6
20 00:04 4,559k 1,027k +0.24 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 g7-g5 Nh4-f3 Be8-g6 g2-g3 a7-a5 g3-g4 h6-h5 h2-h3 b7-b6 Nf3-d2 h5xg4 h3xg4 Bg6-e8 Ke3-f3 Be8-g6 Nd2-c4
21 00:06 9,359k 1,464k +0.24 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 a7-a6 h2-h3 Bh5-d1 b3-b4 Bd1-a4 g2-g4 g7-g5 Nh4-f3 Ba4-e8 Nf3-d2 h6-h5 Nd2-c4 Be8-g6 Nc4xd6 Ke7xd6
22 00:07 12,989k 1,773k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 a7-a6 h2-h3 Bh5-d1 b3-b4 Bd1-a4 g2-g4 g7-g5 Nh4-f3 Ba4-d1 Nf3-d2 Ke7-d7 Nd2-c4 h6-h5 Nc4xd6 Kd7xd6 g4xh5 Bd1xh5 b4-b5 a6xb5 Bd3xb5
23 00:07 16,757k 2,189k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 a7-a6 h2-h3 Bh5-d1 b3-b4 Bd1-a4 g2-g4 g7-g5 Nh4-f3 Ba4-d1 Nf3-d2 Ke7-d7 Nd2-c4 h6-h5 Nc4xd6 Kd7xd6 Bd3-e2 Bd1xe2 Ke3xe2 h5-h4 Ke2-d3 Kd6-e7 Kd3-c4 Ke7-d6 b4-b5 a6xb5+ Kc4xb5
24 00:08 20,184k 2,519k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 a7-a6 h2-h3 Bh5-f7 g2-g4 g7-g5 Nh4-f3 h6-h5 Nf3-d2 h5xg4 h3xg4 Bf7-e8 Bd3-e2 Be8-g6 Nd2-c4 Bg6-e8 Be2-f3 Be8-d7 Nc4-d2
25 00:08 26,868k 3,144k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 Bh5-d1 b3-b4 Bd1-h5 Nh4-f3 Bh5-e8 Nf3-d2 Be8-g6 h2-h3 Bg6-f7 Nd2-f3 Bf7-g6 g2-g4 Bg6-e8 Nf3-h4 g7-g5 Nh4-f5+ Nd6xf5+ g4xf5 a7-a6 Bd3-c4 Ke7-d6 Bc4-e2 Be8-d7
26 00:08 32,694k 3,636k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 Bh5-d1 b3-b4 Bd1-h5 Nh4-f3 Bh5-e8 Nf3-d2 Be8-g6 h2-h3 h6-h5 Nd2-f3 Bg6-e8 Nf3-h4 g7-g5 Nh4-f5+ Nd6xf5+ e4xf5 Ke7-d6
27 00:09 39,268k 4,131k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 Bh5-d1 b3-b4 Bd1-h5 Nh4-f3 Bh5-e8 Nf3-d2 Be8-g6 h2-h3 h6-h5 Nd2-f3 Bg6-e8 a5-a6 b7-b6 Nf3-h4 g7-g5 Nh4-f5+ Nd6xf5+ e4xf5 Ke7-d6 Bd3-e4 Be8-d7
28 00:10 52,325k 5,000k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 Bh5-d1 b3-b4 Bd1-h5 Nh4-f3 Bh5-e8 g2-g3 Be8-d7 Bd3-e2 Bd7-e8 Nf3-h4 g7-g5 Nh4-f3 Be8-f7 h2-h3 Bf7-g6 Be2-d3 h6-h5 Nf3-d2 f6-f5 e4xf5 Nd6xf5+ Ke3-f3 Ke7-d6
29 00:11 59,823k 5,438k +0.20 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 Nf3-h4 Be8-h5 a4-a5 Bh5-d1 b3-b4 Bd1-h5 Nh4-f3 Bh5-e8 g2-g3 Be8-d7 Bd3-e2 Bd7-e8 g3-g4 Be8-g6 Nf3-d2 Bg6-e8 h2-h3 Be8-f7 Be2-d3 Bf7-g6 Nd2-b3 b7-b6 a5xb6 a7xb6 Nb3-d2 h6-h5 Nd2-f3 Bg6-f7 Nf3-h4 h5xg4 h3xg4
30 00:25 267,710k 10,511k +0.16 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Be8-f7 g2-g4 g7-g5 Bc2-d3 Bf7-g6 Nf3-d2 Bg6-e8 Bd3-e2 h5xg4 h3xg4 Be8-d7 Ke3-f3 Bd7-e8 Nd2-c4 Be8-g6 Nc4xd6 Ke7xd6 Kf3-e3 a7-a6 b3-b4 Bg6-e8 b4-b5 a6xb5 Be2xb5 Be8xb5
31 00:32 353,997k 10,974k +0.16 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Be8-f7 g2-g4 g7-g5 Bc2-d3 Bf7-g6 Nf3-d2 Bg6-e8 Bd3-e2 h5xg4 h3xg4 Be8-d7 Ke3-f3 Bd7-e8 Nd2-c4 Be8-g6 Nc4xd6 Ke7xd6 Kf3-e3 Bg6-e8 Be2-d3 a7-a6 Bd3-e2 Be8-g6 b3-b4 Bg6-e8 b4-b5 a6xb5 Be2xb5 Be8xb5 a4xb5 b7-b6 Ke3-d3 Kd6-c5
32 00:36 412,439k 11,200k +0.16 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Be8-f7 g2-g4 g7-g5 Bc2-d3 Bf7-g6 Nf3-d2 Bg6-e8 Bd3-e2 h5xg4 h3xg4 Be8-d7 Ke3-f3 Bd7-e8 Nd2-c4 Be8-g6 Nc4xd6 Ke7xd6 Kf3-e3 Bg6-e8 Be2-d3 a7-a6 Bd3-e2 Be8-d7 Ke3-d3 b7-b5 Kd3-e3 b5xa4 b3xa4 Bd7xa4 Be2xa6
33 00:40 456,951k 11,349k +0.16 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Be8-f7 g2-g4 g7-g5 Bc2-d3 Bf7-g6 Nf3-d2 Bg6-e8 Bd3-e2 h5xg4 h3xg4 Be8-d7 Ke3-f3 Bd7-e8 Nd2-c4 Be8-g6 Nc4xd6 Ke7xd6 Kf3-e3 Bg6-e8 Be2-d3 a7-a6 Bd3-e2 Be8-d7 Ke3-d3 Bd7-e8 Kd3-c4 b7-b5+ Kc4-b4 b5xa4 b3xa4 Be8-d7
34 00:48 558,360k 11,594k +0.16 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-e2 g7-g5 Nh4-f3 h5-h4 Be2-d3 Be8-g6 a4-a5 Bg6-f7 Bd3-c2 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 Bg6-f7 Nd2-c4 Bf7-g6 Bd3-e2 Nd6xe4 Be2-d3
35 00:57 694,244k 11,972k +0.16 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-e2 g7-g5 Nh4-f3 h5-h4 Be2-d3 Be8-g6 a4-a5 Bg6-f7 Bd3-c2 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 Bg6-f7 Nd2-c4 Bf7-g6
36 01:07 828,658k 12,264k +0.16 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-e2 g7-g5 Nh4-f3 h5-h4 Be2-d3 Be8-g6 Nf3-d2 Bg6-f7
37 02:06 1,631,961k 12,929k +0.12 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-c2 Be8-f7 Nh4-f3 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 a7-a6
38 03:20 2,688,731k 13,393k +0.12 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-c2 Be8-f7 Nh4-f3 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 Bg6-e8 h3-h4 Be8-g6 g2-g3 f6-f5 Nd2-f3 f5xe4 Nf3xe5 e4xd3 Ne5xg6+ Ke7-f6 Ng6-f4 Nd6-f5+ Ke3xd3 Nf5xg3 d5-d6 Ng3-f5 d6-d7 Kf6-e7 Kd3-e4 Nf5xh4 Nf4xh5 Ke7xd7 Nh5xg7 Nh4-g6
39 03:42 2,963,686k 13,338k +0.12 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-c2 Be8-f7 Nh4-f3 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 Bg6-e8 h3-h4 Be8-g6 g2-g3 Nd6-f7 Bd3-e2 Nf7-d6 a4-a5 Bg6-e8 Be2-d1 g7-g6 Bd1-e2 Be8-b5 Be2xb5 Nd6xb5 Ke3-d3 Nb5-d6 Nd2-f3 Ke7-d7
40 04:25 3,536,782k 13,337k +0.12 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-c2 Be8-f7 Nh4-f3 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 Bg6-e8 h3-h4 Be8-g6 g2-g3 Nd6-f7 Bd3-e2 Nf7-d6 a4-a5 Bg6-e8 Be2-d1 g7-g6 Bd1-e2 Be8-b5 Be2xb5 Nd6xb5 Ke3-d3 Nb5-d6 Nd2-f3 Ke7-d7 Kd3-e3
41 05:20 4,327,974k 13,523k +0.12 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-c2 Be8-f7 Nh4-f3 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 Bg6-e8 h3-h4 Be8-g6 g2-g3 Nd6-f7 Bd3-e2 Nf7-d6 a4-a5 Bg6-e8 Be2-d1 g7-g6 Bd1-e2 Be8-b5 Be2xb5 Nd6xb5 Ke3-d3 Nb5-d6 Nd2-c4 Nd6xc4
42 05:55 4,768,064k 13,401k +0.12 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-c2 Be8-f7 Nh4-f3 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 Bg6-e8 h3-h4 Be8-g6 g2-g3 Nd6-f7 Bd3-e2 Nf7-d6 a4-a5 Bg6-e8 Be2-d1 g7-g6 Nd2-c4 Be8-d7 Nc4xd6 Ke7xd6 Bd1-e2 b7-b6 a5xb6 a7xb6 Be2-d3
43 07:46 6,310,134k 13,518k +0.12 d4-d5 Nc8-d6 Nd2-f3 Bd7-e8 h2-h3 h6-h5 Bd3-c2 Nd6-f7 Bc2-b1 Nf7-d6 Bb1-d3 Be8-d7 Nf3-h4 Bd7-e8 Bd3-c2 Be8-f7 Nh4-f3 Bf7-e8 Nf3-d2 Be8-g6 Bc2-d3 Bg6-f7
44- 23:31 19,444,726k 13,776k +0.04 d4-d5 Nc8-d6 Nd2-f3 b7-b5 a4-a5 b5-b4 h2-h4 Bd7-e8
44- 51:35 43,511,353k 14,057k -0.04 d4-d5 Ke7-d6
44 53:43 45,408,327k 14,086k 0.00 d4xe5 f6xe5 Bd3-c4 Nc8-d6 Bc4-d5 b7-b5 a4xb5 Nd6xb5 Nd2-c4 Ke7-f6 Ke3-d3 Nb5-d4 Nc4-e3 Kf6-e7 Ne3-c4 Bd7-b5 Kd3-c3
45 1:42:09 86,925,805k 14,181k 0.00 d4xe5 f6xe5 Bd3-c4 Nc8-d6 Bc4-d5 b7-b5 a4xb5 Nd6xb5 Nd2-c4 Ke7-f6 Ke3-d3 Nb5-d4 Nc4-e3 Kf6-e7 Ne3-c4 Bd7-b5 b3-b4
46 1:47:02 91,215,689k 14,202k 0.00 d4xe5 f6xe5 Bd3-c4 Nc8-d6 Bc4-d5 b7-b5 a4xb5 Nd6xb5 Nd2-c4 Ke7-f6 Ke3-d3 Nb5-d4 Nc4-e3 Kf6-e7 Bd5-c4 a7-a5 Bc4-d5 Bd7-b5+ Bd5-c4 Bb5-c6
47 1:53:28 96,936,671k 14,237k 0.00 d4xe5 f6xe5 Bd3-c4 Nc8-d6 Bc4-d5 b7-b5 a4xb5 Nd6xb5 Nd2-c4 Ke7-f6 Ke3-d3 Nb5-d4 Nc4-e3 Kf6-e7 Bd5-c4 a7-a5 Bc4-d5 Bd7-b5+ Bd5-c4 Bb5-c6 Bc4-d5 Bc6xd5 Ne3xd5+ Ke7-d6 Kd3-c4 g7-g6 Nd5-e3 Kd6-e6 Kc4-c3 Nd4-e2+ Kc3-d2 Ne2-d4 Kd2-c3
48 2:01:24 104,105,012k 14,292k 0.00 d4xe5 f6xe5 Bd3-c4 Nc8-d6 Bc4-d5 b7-b5 a4xb5 Nd6xb5 Nd2-c4 Ke7-f6 Ke3-d3 Nb5-d4 Nc4-e3 Kf6-e7 Bd5-c4 a7-a5 Bc4-d5 Bd7-b5+ Bd5-c4 Bb5-c6 Bc4-d5 Bc6xd5 Ne3xd5+ Ke7-d6 Kd3-c4 g7-g6 Nd5-e3 Kd6-e6 Kc4-c3 Nd4-e2+ Kc3-d2 Ne2-d4 Kd2-c3