Kasparov - Topalov, Wijk aan Zee 1999

Discussion of computer chess matches and engine tournaments.

Moderator: Ras

goni-K26
Posts: 312
Joined: Mon Mar 09, 2026 4:16 pm
Full name: Agron Kovaci

Kasparov - Topalov, Wijk aan Zee 1999

Post by goni-K26 »

Kasparov - Topalov, Wijk aan Zee 1999: do engines find the Immortal?

Hi all,

everybody here knows this game. Kasparov plays 24.Rxd4!!, gives up a whole rook, and then chases Topalov's king from b6 all the way down to e1. Many people still call it the most beautiful game ever played.

For a long time I wanted to see how today's engines deal with it. Not by analysing it with one engine, but by letting a big group of engines play the position against each other, many times, on both sides of the board. Two simple questions:
  • With White: does the engine find 24.Rxd4, or does it play something "safer"?
  • With Black: does it take the rook like Topalov did (24...cxd4), or does it decline?
The idea

The opening is forced up to 23...Qd6 (46 plies, exactly the moves of the real game). From move 24 on, the engines are on their own. So every game starts at the critical moment, with White to move.

To compare each game with the original I wrote a small tool (PGN Line Matcher). For every game it gives two numbers:
  • Same = how many plies are identical to Kasparov-Topalov, move by move.
  • Pos = the deepest point of the original game that the engine game reaches as a position, even if the moves came in a different order (transpositions).
The original game has 87 plies. The book ends at ply 46, so every game has Same >= 46.

Conditions

Code: Select all

 +----------------------+----------------------------------------+
| Feature              | Technical Specifications               |
+----------------------+----------------------------------------+
| Tournament Series    | Kasparov - Topalov, Wijk aan Zee 1999  |
| Format & Logic       | Round-Robin                            |
| Time Control         | 7s + 0.1s (from move 24)               |
| Main Focus           | Speed-Efficiency & Engine Evolution    |
+----------------------+----------------------------------------+
| Rounds (Games)       | 8 (3480 Games)                         |
| Hash Size            | 16 MB (Optimized for Ultra-Bullet)     |
| CPU Threads          | Single Thread (Pure Core Performance)  |
| Endgame Support      | 5-man Syzygy Tablebases                |
+----------------------+----------------------------------------+
| Interface / GUI      | CuteChess-Cli                          |
| Processor (CPU)      | Intel i5-4570 3.20GHz 8 GB RAM         |
| System Environment   | Windows 11 Pro x64                     |
+----------------------+----------------------------------------+
About the time control: most of my tests run at 9s+0.1, which is 5% of a standard 3m+2s blitz game. Here the game starts at move 24, so about a third of it is already "played", and I took 2 seconds off.

One more thing. I have a rule for quick draws: if a pair produces more than 3 quick draws, the game is replayed from a normal opening. This happened 26 times. Those games count in the standings, but not in the opening statistics below, because they don't start from the Kasparov-Topalov position.

Participants (Elo from my own rating list)

Code: Select all

  1  Reckless 0.9.0      2857     16  Motor 0.9.0         2756
  2  PlentyChess 8.0.0   2844     17  Minke 7.0.0         2750
  3  Caissa 2.0.0        2820     18  Horsie 1.1          2749
  4  Alexandria 8.1.12   2820     19  Tarnished 5.0       2742
  5  Viridithas 20.0.0   2817     20  Igel 3.8.0          2740
  6  Berserk 14.0        2806     21  Tcheran 13.0        2730
  7  Obsidian 16.0       2798     22  Halogen 16.0        2719
  8  Cinder 0.6.1        2790     23  Koivisto 9.0        2713
  9  Integral 8.0        2782     24  Velvet 8.1.1        2712
 10  Devre 7.0           2780     25  Clarity 8.0.0       2706
 11  RubiChess 20240817  2779     26  Revenge 4.0         2698
 12  Stormphrax 8.0.0    2775     27  RofChade 3.1        2682
 13  Komodo Dragon 3.3   2771     28  Booot 7.4           2681
 14  Clover 9.1          2768     29  Rebel 16.3          2660
 15  Starzix 6.0         2757     30  Patricia 5.1.0      2648
In the next posts I'll go through the test two rounds at a time: what the engines played at move 24, where they left the real game, which games came closest to the original, and the standings.

Spoiler for the curious: move 24 decides almost everything in this test, much more than Elo does.
Top
goni-K26
Posts: 312
Joined: Mon Mar 09, 2026 4:16 pm
Full name: Agron Kovaci

Kasparov - Topalov, Wijk aan Zee 1999(Rounds 1-2)

Post by goni-K26 »

Kasparov - Topalov test: rounds 1-2

The first two rounds (870 games, one of them replayed from a normal opening because of the quick-draw rule) already give a very clear picture. After 23...Qd6 the engines chose only two moves with White: 24.Rxd4 or 24.Nc6+. Nothing else.

Code: Select all

 White's 24th / Black's reply   Games    1-0    1/2    0-1   White score
 ----------------------------   -----   ----   ----   ----   -----------
 24.Nc6+                          285      5     74    206           15%
 24.Rxd4 Kb6 (declined)           457      3    425     29           47%
 24.Rxd4 cxd4 (accepted)          127     69     50      8           74%
That table says a lot:
  • 24.Nc6+ is close to a disaster for White: 15%.
  • After 24.Rxd4 the correct answer is 24...Kb6. With it Black keeps the balance: 425 of 457 games ended in a draw.
  • Taking the rook like Topalov (24...cxd4) costs Black a lot: White scores 74%.
Who is Kasparov, who is Topalov?

The engines are very consistent here. Almost every engine is either 0% or close to 100%, on both sides of the board:
  • Kasparov group (find 24.Rxd4 and never take the rook): Berserk, Booot, Clover, Halogen, Horsie, Minke, Obsidian, PlentyChess, Rebel, Reckless, Revenge, Tarnished, Velvet, Stormphrax, Cinder, Igel, Viridithas, Alexandria.
  • Topalov group (never play 24.Rxd4, always take the rook): Komodo Dragon 3.3, Integral 8.0, RofChade 3.1, Clarity 8.0.0.
  • The strange ones: Caissa 2.0.0 and RubiChess play 24.Rxd4 every time with White, but with Black they take the rook (18/19 and 19/20).
  • Defend but don't attack: Devre, Patricia, Starzix and Tcheran never play 24.Rxd4, but they also never fall for 24...cxd4.
  • Undecided: Koivisto (Rxd4 10/29) and Motor (cxd4 5/20).
Seeing Komodo Dragon in the Topalov group surprised me. It's a strong engine (2771 in my list), but in this position it fails on both sides.

Where did the games leave Kasparov-Topalov?

Only the 127 games where Black took the rook can follow the real game. This is where they left it:

Code: Select all

 Original move      Played instead    Games   White score
 ----------------   ---------------   -----   -----------
 25. Re7+           25. Qxd4+             8            0%
 28. Qc3            28. Ra7              16          100%
 30... Qc4          30... Rd6            25           96%
 30... Qc4          30... Rhe8           11           86%
 31... Kxa3         31... Rd1+           18           58%
 33. c3+            33. Bd7              34           56%
 35... Kd1          35... Ke1             5          100%
 39... Rd3          39... Kd2             3          100%
 39... Rd3          39... f5              1          100%
 40. Qa8            40. Qa1               1          100%
 40. Qa8            40. Qc8               2          100%
 40... c3           40... Rb3+            1          100%
 40... c3           40... f5              2          100%
Some things worth noticing:
  • 25.Qxd4+ instead of Kasparov's 25.Re7+ loses every time. Most of these games were played by Caissa with White.
  • 28.Ra7 instead of 28.Qc3, mostly by PlentyChess and Tarnished: 16 games, 16 wins.
  • 33.c3+ is the move that matters. Engines that played 33.Bd7 instead got mostly draws (56%).
  • 31...Rd1+ is a much better defence than Topalov's 31...Kxa3. Integral played it 16 times out of 18.
Closest to the original

3 games reached Kasparov's 40.Qa8 (Same 79). In all three, Black then played something other than Topalov's 40...c3 (40...Rb3+ or 40...f5).

There were also two transpositions: in Cinder - Komodo Dragon and Clover - Komodo Dragon, Black played 35...Ke1 instead of 35...Kd1, and after 36.Re7+ Kd1 37.Bf1 Rd2 38.Rd7 the game was back in the original position (Pos 78).

Standings after round 2

Code: Select all

 Rank  Engine               Elo   Points  Games  Score   Elo rank
 ----  -------------------  ----  ------  -----  ------  --------
    1  PlentyChess 8.0.0    2844    37.5     58   64.7%         2
    2  Obsidian 16.0        2798    37.0     58   63.8%         7
    3  Reckless 0.9.0       2857    36.5     58   62.9%         1
    4  Berserk 14.0         2806    36.0     58   62.1%         6
    5  Clover 9.1           2768    35.5     58   61.2%        14
    6  Tarnished 5.0        2742    35.5     58   61.2%        19
    7  Cinder 0.6.1         2790    35.0     58   60.3%         8
    8  Halogen 16.0         2719    35.0     58   60.3%        22
    9  Stormphrax 8.0.0     2775    35.0     58   60.3%        12
   10  Viridithas 20.0.0    2817    35.0     58   60.3%         5
   11  Minke 7.0.0          2750    34.5     58   59.5%        17
   12  Velvet 8.1.1         2712    34.5     58   59.5%        24
   13  Horsie 1.1           2749    34.5     58   59.5%        18
   14  Alexandria 8.1.12    2820    33.5     58   57.8%         3
   15  Igel 3.8.0           2740    33.5     58   57.8%        20
   16  Revenge 4.0          2698    33.0     58   56.9%        26
   17  Booot 7.4            2681    31.0     58   53.4%        28
   18  RubiChess 20240817   2779    30.5     58   52.6%        11
   19  Caissa 2.0.0         2820    26.0     58   44.8%         3
   20  Devre 7.0            2780    25.0     58   43.1%        10
   21  Koivisto 9.0         2713    24.5     58   42.2%        23
   22  Tcheran 13.0         2730    24.0     58   41.4%        21
   23  Integral 8.0         2782    23.5     58   40.5%         9
   24  Rebel 16.3           2660    23.0     58   39.7%        29
   25  Komodo Dragon 3.3    2771    22.0     58   37.9%        13
   26  Motor 0.9.0          2756    21.5     58   37.1%        16
   27  Starzix 6.0          2757    21.0     58   36.2%        15
   28  Patricia 5.1.0       2648    17.0     58   29.3%        30
   29  RofChade 3.1         2682    10.5     58   18.1%        27
   30  Clarity 8.0.0        2706     9.0     58   15.5%        25
Places 20 to 30 are taken by the engines that never play 24.Rxd4, plus Koivisto and Rebel. Caissa (Elo rank 3) is only 19th.

Horsie 1.1 - Caissa 2.0.0 (round 1, 1-0) The first game in the test to follow Kasparov-Topalov all the way to 40.Qa8. Here Caissa goes for 40...Rb3+ instead of Topalov’s 40...c3, but the ending is already lost: Horsie collects the c-pawn, then the f- and h-pawns, and finishes with a clean mate, 60.Qc2#.

[pgn][Event "Kasparov-Topalov"]
[Site "Intel(R) Core(TM) i5-4570 3.20GHz 8GB RAM"]
[Date "2026.10.31"]
[Round "1"]
[White "Horsie 1.1 x64(AVX2-5Pct)"]
[Black "Caissa 2.0.0 x64(BMI2-5Pct)"]
[Result "1-0"]
[BlackElo "2820"]
[ECO "B07"]
[Opening "Pirc defense"]
[Time "15:50:30"]
[Duration "00:00:17"]
[WhiteElo "2749"]
[TimeControl "7+0.1"]
[Termination "normal"]
[PlyCount "119"]

1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 Bg7 5. Qd2 c6 6. f3 b5 7. Nge2 Nbd7 8. Bh6
Bxh6 9. Qxh6 Bb7 10. a3 e5 11. O-O-O Qe7 12. Kb1 a6 13. Nc1 O-O-O 14. Nb3 exd4
15. Rxd4 c5 16. Rd1 Nb6 17. g3 Kb8 18. Na5 Ba8 19. Bh3 d5 20. Qf4+ Ka7 21. Rhe1
d4 22. Nd5 Nbxd5 23. exd5 Qd6 24. Rxd4 {-0.48/15 0.55} cxd4 {+1.86/16 0.63} 25.
Re7+ {+1.03/16 1.10} Kb6 {+2.16/12 0.08} 26. Qxd4+ {+0.82/19 0.31} Kxa5
{+1.61/10 0.07} 27. b4+ {+0.81/19 0.17} Ka4 {+1.12/13 0.09} 28. Qc3
{+0.83/20 0.23} Qxd5 {-0.58/13 0.11} 29. Ra7 {+0.90/18 0.20} Bb7
{-0.04/13 0.07} 30. Rxb7 {+0.94/20 0.24} Qc4 {-0.27/16 0.17} 31. Qxf6
{+0.92/20 0.21} Kxa3 {-0.51/21 0.95} 32. Qxa6+ {+2.90/19 0.15} Kxb4 33. c3+
{+3.29/17 0.16} Kxc3 {-2.61/15 0.23} 34. Qa1+ {+3.25/17 0.17} Kd2
{-3.31/16 0.34} 35. Qb2+ {+3.34/16 0.15} Kd1 {-3.11/21 0.75} 36. Bf1
{+3.41/16 0.15} Rd2 {-3.40/17 0.28} 37. Rd7 {+3.51/15 0.21} Rxd7
{-3.50/11 0.06} 38. Bxc4 {+3.51/16 0.22} bxc4 {-3.55/14 0.06} 39. Qxh8
{+3.35/16 0.17} Rd3 {-3.58/18 0.48} 40. Qa8 {+3.48/14 0.16} Rb3+
{-3.86/19 1.10} 41. Ka2 {+3.88/15 0.15} Rd3 {-4.29/19 0.55} 42. Qa4+
{+3.99/15 0.14} Ke1 {-4.47/15 0.25} 43. Qxc4 {+4.16/15 0.20} Rxf3
{-4.30/13 0.14} 44. g4 {+4.17/16 0.22} Rf2+ {-3.89/18 0.63} 45. Kb1
{+4.48/15 0.25} Rf1 {-4.67/18 0.76} 46. Kc2 {+4.83/15 0.19} Rf2+
{-4.83/18 0.33} 47. Kd3 {+4.99/15 0.13} Rxh2 {-4.83/16 0.15} 48. Qxf7
{+5.64/15 0.18} Rh3+ {-5.00/16 0.15} 49. Kd4 {+5.77/14 0.12} Ke2
{-5.64/19 0.39} 50. Qf4 {+5.98/16 0.18} g5 {-6.27/19 0.65} 51. Qxg5
{+7.03/16 0.15} Kf2 {-7.02/16 0.23} 52. Qf4+ {+8.30/17 0.19} Kg2
{-8.06/13 0.13} 53. g5 {+8.89/17 0.18} Rb3 {-306.96/24 0.11} 54. Qe4+
{+9.66/14 0.15} Kg3 {-306.98/26 0.10} 55. Qxh7 {+10.02/16 0.19} Rb4+ 56. Kc3
{+10.21/16 0.17} Kf2 57. Kxb4 {+15.78/16 0.11} Ke1 58. Qh2 {+M7/63 0.26} Kd1
59. Kc3 {+M5/63 0.02} Kc1 60. Qc2# {+M1/63 0.01 White mates} 1-0[/pgn]

Viridithas 20.0.0 - RubiChess (round 1, 1-0) Another game that is identical to the real one up to 40.Qa8. This time Black tries 40...f5 and runs with the king towards the kingside. Viridithas stays calm, wins the f-pawn, swaps queen for rook at the right moment and queens the h-pawn. The game was adjudicated at move 66.
[pgn][Event "Kasparov-Topalov"]
[Site "Intel(R) Core(TM) i5-4570 3.20GHz 8GB RAM"]
[Date "2026.10.31"]
[Round "1"]
[White "Viridithas 20.0.0 x64(AVX2-5Pct)"]
[Black "RubiChess 20240817 x64(BMI2-5Pct)"]
[Result "1-0"]
[BlackElo "2779"]
[ECO "B07"]
[Opening "Pirc defense"]
[Time "17:37:15"]
[Duration "00:00:16"]
[WhiteElo "2817"]
[TimeControl "7+0.1"]
[Termination "adjudication"]
[PlyCount "132"]

1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 Bg7 5. Qd2 c6 6. f3 b5 7. Nge2 Nbd7 8. Bh6
Bxh6 9. Qxh6 Bb7 10. a3 e5 11. O-O-O Qe7 12. Kb1 a6 13. Nc1 O-O-O 14. Nb3 exd4
15. Rxd4 c5 16. Rd1 Nb6 17. g3 Kb8 18. Na5 Ba8 19. Bh3 d5 20. Qf4+ Ka7 21. Rhe1
d4 22. Nd5 Nbxd5 23. exd5 Qd6 24. Rxd4 {-1.05/20 1.20} cxd4 {+1.33/19 0.85} 25.
Re7+ {+1.47/22 0.19} Kb6 {-1.01/17 0.31} 26. Qxd4+ {+1.33/21 0.19} Kxa5
{-0.94/18 0.18} 27. b4+ {+2.76/20 0.18} Ka4 {-0.94/21 0.15} 28. Qc3
{+1.29/22 0.31} Qxd5 {-0.79/20 0.16} 29. Ra7 {+1.51/23 0.22} Bb7
{-0.83/20 0.18} 30. Rxb7 {+1.15/21 0.27} Qc4 {-1.07/19 0.35} 31. Qxf6
{+1.54/22 0.25} Kxa3 {-1.00/17 0.16} 32. Qxa6+ {+6.88/25 0.24} Kxb4
{-1.00/5 0.00} 33. c3+ {+6.53/20 0.14} Kxc3 {-1.81/19 0.39} 34. Qa1+
{+6.78/16 0.08} Kd2 {-2.00/19 0.24} 35. Qb2+ {+7.08/15 0.06} Kd1
{-1.78/19 0.41} 36. Bf1 {+7.60/13 0.09} Rd2 {-2.33/19 0.47} 37. Rd7
{+7.60/16 0.09} Rxd7 {-2.35/18 0.24} 38. Bxc4 {+8.36/15 0.16} bxc4
{-2.19/18 0.23} 39. Qxh8 {+8.70/16 0.22} Rd3 {-2.45/19 1.30} 40. Qa8
{+9.13/15 0.21} f5 {-2.95/18 2.10} 41. Qc6 {+9.52/16 0.18} Ke1 {-2.91/17 1.10}
42. g4 {+9.90/15 0.17} Kd2 {-2.99/13 0.10} 43. gxf5 {+11.18/16 0.16} gxf5
{-3.28/13 0.10} 44. Qxc4 {+11.92/15 0.09} Rxf3 {-3.38/15 0.10} 45. Qb4+
{+12.33/16 0.11} Ke2 {-3.41/13 0.10} 46. Qe7+ {+12.96/17 0.12} Kf2
{-3.76/13 0.10} 47. Qxh7 {+14.87/14 0.10} Kg2 {-4.30/13 0.10} 48. Kc2
{+16.31/14 0.08} Rf2+ {-4.23/12 0.10} 49. Kd3 {+19.85/14 0.09} f4
{-4.57/12 0.10} 50. h4 {+18.21/15 0.06} f3 {-5.02/13 0.10} 51. Qg6+
{+299.93/28 0.11} Kh1 {-129.99/17 0.10} 52. Qg4 {+299.95/25 0.08} Kh2
{-129.99/17 0.10} 53. Ke3 {+299.97/25 0.07} Ra2 {-130.00/17 0.10} 54. Kxf3
{+299.99/18 0.05} Ra3+ {-130.01/18 0.10} 55. Kf2 {+14.41/18 0.00} Ra2+
{-130.02/19 0.10} 56. Qe2 {+14.41/18 0.00} Rxe2+ {-6.98/22 0.10} 57. Kxe2
{+14.41/18 0.00} Kg1 {-130.02/19 0.10} 58. h5 {+14.41/18 0.00} Kg2
{-8.15/20 0.10} 59. h6 {+14.41/18 0.00} Kg3 {-8.72/20 0.10} 60. h7
{+14.41/18 0.00} Kf4 {-8.65/21 0.10} 61. h8=Q {+14.41/18 0.00} Ke4
{-M26/17 0.10} 62. Qd8 {+14.41/18 0.00} Ke5 {-28.99/16 0.10} 63. Ke3
{+14.41/18 0.00} Kf5 {-M10/15 0.04} 64. Qd6 {+14.41/18 0.00} Kg5 {-M10/10 0.00}
65. Ke4 {+14.41/18 0.00} Kh5 {-M8/9 0.00} 66. Kf4 {+14.41/18 0.00} Kh4
{-M4/5 0.00 White wins by adjudication} 1-0
[/pgn]

Cinder 0.6.1 - RubiChess (round 2, 1-0) The same line again, and again 40...f5, but Cinder takes a different road: 41.Qa4+ Ke1 42.Qxc4, and then it simply pushes the h-pawn. RubiChess gives a few checks with the rook, but 55.h8=Q ends the story.
[pgn][Event "Kasparov-Topalov"]
[Site "Intel(R) Core(TM) i5-4570 3.20GHz 8GB RAM"]
[Date "2026.10.31"]
[Round "2"]
[White "Cinder 0.6.1 x64(AVX2-5Pct)"]
[Black "RubiChess 20240817 x64(BMI2-5Pct)"]
[Result "1-0"]
[BlackElo "2779"]
[ECO "B07"]
[Opening "Pirc defense"]
[Time "19:03:44"]
[Duration "00:00:18"]
[WhiteElo "2790"]
[TimeControl "7+0.1"]
[Termination "adjudication"]
[PlyCount "118"]

1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 Bg7 5. Qd2 c6 6. f3 b5 7. Nge2 Nbd7 8. Bh6
Bxh6 9. Qxh6 Bb7 10. a3 e5 11. O-O-O Qe7 12. Kb1 a6 13. Nc1 O-O-O 14. Nb3 exd4
15. Rxd4 c5 16. Rd1 Nb6 17. g3 Kb8 18. Na5 Ba8 19. Bh3 d5 20. Qf4+ Ka7 21. Rhe1
d4 22. Nd5 Nbxd5 23. exd5 Qd6 24. Rxd4 {-0.35/18 1.20} cxd4 {+1.33/19 0.78} 25.
Re7+ {+1.19/15 0.14} Kb6 {-1.01/17 0.28} 26. Qxd4+ {+0.53/20 0.42} Kxa5
{-0.94/18 0.18} 27. b4+ {+0.92/20 0.28} Ka4 {-0.94/21 0.14} 28. Qc3
{+0.16/17 0.77} Qxd5 {-0.79/20 0.15} 29. Ra7 {+0.07/18 0.48} Bb7
{-0.83/20 0.16} 30. Rxb7 {+0.41/16 0.29} Qc4 {-1.07/19 0.33} 31. Qxf6
{+1.21/15 0.23} Kxa3 {-1.00/17 0.16} 32. Qxa6+ {+2.57/16 0.11} Kxb4
{-1.00/5 0.00} 33. c3+ {+2.96/16 0.14} Kxc3 {-1.87/20 0.66} 34. Qa1+
{+3.06/17 0.11} Kd2 {-1.96/20 0.28} 35. Qb2+ {+2.98/17 0.18} Kd1
{-1.89/22 0.43} 36. Bf1 {+3.11/16 0.10} Rd2 {-2.08/19 0.28} 37. Rd7
{+3.20/15 0.18} Rxd7 {-2.08/18 0.38} 38. Bxc4 {+3.14/16 0.17} bxc4
{-2.20/17 0.38} 39. Qxh8 {+3.07/14 0.16} Rd3 {-2.88/18 1.30} 40. Qa8
{+3.16/15 0.17} f5 {-3.05/18 2.70} 41. Qa4+ {+3.34/15 0.22} Ke1 {-2.85/14 0.10}
42. Qxc4 {+3.47/13 0.14} Rxf3 {-2.83/14 0.10} 43. Qh4 {+3.55/13 0.16} Kf1
{-3.04/13 0.10} 44. Qxh7 {+3.83/14 0.15} g5 {-2.67/14 0.10} 45. h4
{+4.20/13 0.13} g4 {-4.07/14 0.10} 46. Qd7 {+4.39/12 0.15} Kg1 {-4.30/12 0.10}
47. h5 {+4.56/13 0.14} Rf1+ {-4.78/14 0.10} 48. Kc2 {+5.19/12 0.12} Rf2+
{-5.12/13 0.10} 49. Kd3 {+5.41/12 0.11} Rf3+ {-5.35/12 0.10} 50. Kd4
{+5.92/12 0.10} Rf2 {-5.35/12 0.10} 51. Qg7 {+6.82/12 0.10} Re2 {-4.98/13 0.10}
52. h6 {+8.50/13 0.08} Re8 {-5.79/12 0.10} 53. h7 {+11.85/14 0.07} Re4+
{-7.02/13 0.10} 54. Kd3 {+15.97/14 0.06} Re1 {-11.33/14 0.10} 55. h8=Q
{+68.92/13 0.04} Rd1+ {-129.97/16 0.10} 56. Ke2 {+102.85/14 0.09} Rf1
{-M12/13 0.02} 57. Qgh6 {+M13/18 0.14} Rf2+ {-M14/14 0.02} 58. Kd3
{+M11/23 0.08} Rf3+ {-M10/12 0.00} 59. Kc4 {+M9/42 0.50} Ra3
{-M8/8 0.00 White wins by adjudication} 1-0
[/pgn]

Your turn: after two rounds, 4 engines never play 24.Rxd4 and always take the rook as Black. Do you think any of them will “wake up” in the next 6 rounds, or is this a blind spot that time and luck can’t fix? Place your bets before the next post!
Top
Vladimir Xern
Posts: 42
Joined: Wed Mar 08, 2006 8:30 pm

Re: Kasparov - Topalov, Wijk aan Zee 1999

Post by Vladimir Xern »

What a fun, well-presented experiment.

I'm both surprised how far some engines followed the human precedent and curious how many other "immortal" games stand up to this kind of scattershot sampling.
Top
goni-K26
Posts: 312
Joined: Mon Mar 09, 2026 4:16 pm
Full name: Agron Kovaci

Kasparov - Topalov, Wijk aan Zee 1999(Rounds 3-4)

Post by goni-K26 »

Kasparov - Topalov test: rounds 3-4

Rounds 3 and 4 (865 games) were almost a copy of rounds 1-2. That is a result in itself: in this position the engines are very deterministic.

Code: Select all

 White's 24th / Black's reply   Games    1-0    1/2    0-1   White score
 ----------------------------   -----   ----   ----   ----   -----------
 24.Nc6+                          277      5     81    191           16%
 24.Rxd4 Kb6 (declined)           458      1    433     24           47%
 24.Rxd4 cxd4 (accepted)          130     73     50      7           75%
The groups didn't change. Komodo Dragon, Integral, RofChade and Clarity again took the rook every time it was offered. With White, only Clarity "woke up" a little: 24.Rxd4 in 2 of its 29 games. Caissa (20/20) and RubiChess (20/21) did the same. Koivisto played 24.Rxd4 a bit more often (13/29), and Motor took the rook 8 times out of 20.

Code: Select all

 Original move      Played instead    Games   White score
 ----------------   ---------------   -----   -----------
 25. Re7+           25. Qxd4+             5            0%
 27. b4+            27. Qc3+              2            0%
 28. Qc3            28. Ra7               9          100%
 30... Qc4          30... Rd6            31           95%
 30... Qc4          30... Rhe8            6          100%
 31... Kxa3         31... Rd1+           20           62%
 33. c3+            33. Bd7              38           58%
 39... Rd3          39... Kd2             1          100%
 39... Rd3          39... f5              1          100%
 40. Qa8            40. Qa1               5          100%
 40. Qa8            40. Qc8               5          100%
 40... c3           40... Rb3+            1          100%
 40... c3           40... f5              5          100%
 40... c3           40... g5              1          100%
Komodo Dragon, the real Topalov

The nice part of these rounds: 7 games reached 40.Qa8 (Same 79), and in 5 of them Black was Komodo Dragon. RubiChess had the other 2. So far, these two engines follow Topalov's moves longer than anybody else.

Black's 40th move after 40.Qa8 was 40...f5 (5 times), 40...Rb3+ and 40...g5. Still no 40...c3.

Some games left the real game one move earlier, at White's 40th move:
  • Reckless plays 40.Qc8 instead of 40.Qa8, every time it got there so far. Horsie tried it twice too.
  • Cinder, Stormphrax and Horsie sometimes play 40.Qa1.
Three games worth a look:

Horsie 1.1 - Komodo Dragon 3.3 (round 4, 1-0): the only game so far with 40...g5 after 40.Qa8.
[pgn]
[Event "Kasparov-Topalov"]
[Site "Intel(R) Core(TM) i5-4570 3.20GHz 8GB RAM"]
[Date "2026.11.01"]
[Round "4"]
[White "Horsie 1.1 x64(AVX2-5Pct)"]
[Black "Komodo Dragon 3.3 x64(AVX2-5Pct)"]
[Result "1-0"]
[BlackElo "2771"]
[ECO "B07"]
[Opening "Pirc defense"]
[Time "00:27:34"]
[Duration "00:00:16"]
[WhiteElo "2749"]
[TimeControl "7+0.1"]
[Termination "adjudication"]
[PlyCount "114"]

1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 Bg7 5. Qd2 c6 6. f3 b5 7. Nge2 Nbd7 8. Bh6
Bxh6 9. Qxh6 Bb7 10. a3 e5 11. O-O-O Qe7 12. Kb1 a6 13. Nc1 O-O-O 14. Nb3 exd4
15. Rxd4 c5 16. Rd1 Nb6 17. g3 Kb8 18. Na5 Ba8 19. Bh3 d5 20. Qf4+ Ka7 21. Rhe1
d4 22. Nd5 Nbxd5 23. exd5 Qd6 24. Rxd4 {-0.48/15 0.48} cxd4 {+2.01/15 0.38} 25.
Re7+ {+1.03/16 1.20} Kb6 {+2.11/18 0.17} 26. Qxd4+ {+0.82/19 0.26} Kxa5
{+2.28/19 0.28} 27. b4+ {+0.87/20 0.18} Ka4 {-0.54/17 0.24} 28. Qc3
{+0.83/19 0.28} Qxd5 {-0.42/20 0.25} 29. Ra7 {+0.78/20 0.27} Bb7
{-0.42/19 0.23} 30. Rxb7 {+0.90/20 0.26} Qc4 {-0.37/17 0.26} 31. Qxf6
{+1.06/18 0.15} Kxa3 {-0.36/19 0.52} 32. Qxa6+ {+3.15/21 0.32} Kxb4
{-0.36/5 0.00} 33. c3+ {+3.16/19 0.21} Kxc3 {-1.54/15 0.16} 34. Qa1+
{+3.41/18 0.18} Kd2 {-3.19/17 0.22} 35. Qb2+ {+3.25/17 0.14} Kd1
{-3.37/17 0.24} 36. Bf1 {+3.48/15 0.30} Rd2 {-3.12/15 0.13} 37. Rd7
{+3.57/17 0.37} Rxd7 {-3.36/16 0.15} 38. Bxc4 {+3.81/18 0.48} bxc4
{-3.45/16 0.17} 39. Qxh8 {+3.63/16 0.14} Rd3 {-3.42/14 0.11} 40. Qa8
{+3.79/16 0.22} g5 {-3.69/16 0.64} 41. Qe4 {+5.33/14 0.17} Rb3+ {-4.05/14 0.24}
42. Ka2 {+5.29/13 0.12} f5 {-4.47/14 0.31} 43. Qxf5 {+6.07/14 0.16} Rd3
{-4.67/14 0.47} 44. Qb5 {+6.35/13 0.11} Rd4 {-4.93/13 0.41} 45. Kb1
{+7.28/15 0.15} h6 {-4.71/13 0.28} 46. f4 {+8.34/15 0.13} Kd2 {-6.53/13 0.26}
47. f5 {+9.63/17 0.23} c3 {-6.19/13 0.26} 48. Qc5 {+9.97/15 0.17} Rd8
{-7.52/13 0.37} 49. Qf2+ {+10.35/14 0.14} Kd3 {-7.42/12 0.09} 50. Qc2+
{+10.56/16 0.20} Ke3 {-5.92/12 0.12} 51. Qxc3+ {+10.54/15 0.12} Ke4
{-7.68/12 0.13} 52. f6 {+11.70/14 0.28} Rh8 {-10.51/12 0.21} 53. f7
{+13.21/14 0.40} Rb8+ {-11.94/13 0.15} 54. Kc2 {+13.19/12 0.09} Kd5
{-13.62/14 0.14} 55. Qc7 {+13.33/12 0.14} Rf8 {-17.88/15 0.18} 56. Qd7+
{+13.35/10 0.09} Ke4 {-250.00/13 0.02} 57. Qe7+ {+14.17/12 0.32} Kf3
{-M16/14 0.01 White wins by adjudication} 1-0
[/pgn]


Reckless 0.9.0 - Komodo Dragon 3.3 (round 4, 1-0): Reckless leaves the real game with 40.Qc8 instead of 40.Qa8. Dragon pushes the c-pawn anyway (40...c3), but with the queen on c8 it is a different position.

[pgn]
[Event "Kasparov-Topalov"]
[Site "Intel(R) Core(TM) i5-4570 3.20GHz 8GB RAM"]
[Date "2026.11.01"]
[Round "4"]
[White "Reckless 0.9.0 x64(AVX2-5Pct)"]
[Black "Komodo Dragon 3.3 x64(AVX2-5Pct)"]
[Result "1-0"]
[BlackElo "2771"]
[ECO "B07"]
[Opening "Pirc defense"]
[Time "02:13:36"]
[Duration "00:00:21"]
[WhiteElo "2880"]
[TimeControl "7+0.1"]
[Termination "adjudication"]
[PlyCount "140"]

1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 Bg7 5. Qd2 c6 6. f3 b5 7. Nge2 Nbd7 8. Bh6
Bxh6 9. Qxh6 Bb7 10. a3 e5 11. O-O-O Qe7 12. Kb1 a6 13. Nc1 O-O-O 14. Nb3 exd4
15. Rxd4 c5 16. Rd1 Nb6 17. g3 Kb8 18. Na5 Ba8 19. Bh3 d5 20. Qf4+ Ka7 21. Rhe1
d4 22. Nd5 Nbxd5 23. exd5 Qd6 24. Rxd4 {-0.37/20 1.70} cxd4 {+2.13/18 0.57} 25.
Re7+ {+1.17/15 0.24} Kb6 {+2.04/18 0.22} 26. Qxd4+ {+1.43/16 0.24} Kxa5
{+2.08/19 0.23} 27. b4+ {+1.48/16 0.26} Ka4 {+0.81/15 0.42} 28. Qc3
{+1.72/20 0.90} Qxd5 {-0.36/20 0.45} 29. Ra7 {+1.78/17 0.22} Bb7
{-0.43/20 0.27} 30. Rxb7 {+1.23/16 0.25} Qc4 {-0.48/18 0.19} 31. Qxf6
{+1.40/16 0.15} Kxa3 {-0.37/18 0.29} 32. Qxa6+ {+3.72/16 0.17} Kxb4
{-0.37/5 0.00} 33. c3+ {+3.60/14 0.24} Kxc3 {-2.11/14 0.13} 34. Qa1+
{+3.68/11 0.13} Kd2 {-2.37/16 0.13} 35. Qb2+ {+3.76/12 0.23} Kd1
{-3.01/16 0.14} 36. Bf1 {+3.94/13 0.14} Rd2 {-3.27/16 0.18} 37. Rd7
{+3.92/14 0.25} Rxd7 {-3.40/15 0.13} 38. Bxc4 {+4.54/14 0.39} bxc4
{-3.55/15 0.11} 39. Qxh8 {+4.86/13 0.14} Rd3 {-3.46/14 0.17} 40. Qc8
{+4.89/12 0.20} c3 {-3.22/15 0.28} 41. f4 {+5.65/11 0.15} f5 {-3.79/14 0.28}
42. Qc4 {+5.73/12 0.17} Kd2 {-3.92/13 0.16} 43. Qa2+ {+5.98/11 0.12} Ke1
{-4.04/13 0.10} 44. Qa7 {+5.99/13 0.21} Kd1 {-4.30/15 0.50} 45. Qg1+
{+6.09/13 0.13} Ke2 {-4.54/13 0.14} 46. Kc2 {+6.09/11 0.08} Rd2+
{-4.50/14 0.32} 47. Kxc3 {+6.22/13 0.12} Rd3+ {-5.36/13 0.13} 48. Kb4
{+6.17/14 0.27} Rd1 {-5.39/13 0.46} 49. Qa7 {+6.02/16 0.25} Rb1+
{-6.26/13 0.68} 50. Kc3 {+6.02/15 0.08} Rc1+ {-6.90/13 0.35} 51. Kd4
{+6.24/13 0.14} Rc6 {-7.23/13 0.51} 52. Ke5 {+6.24/14 0.14} h6 {-7.12/12 0.19}
53. Qb7 {+6.47/12 0.14} Rc2 {-8.03/10 0.17} 54. Kf6 {+6.42/15 0.46} Rd2
{-8.20/10 0.19} 55. Kxg6 {+6.44/16 0.36} h5 {-8.59/10 0.24} 56. Qb4
{+6.43/13 0.11} Rc2 {-8.35/10 0.24} 57. Qb6 {+6.43/14 0.20} Rd2 {-8.98/11 0.48}
58. Kg5 {+6.35/17 0.58} Kf1 {-8.54/10 0.13} 59. h3 {+6.35/13 0.11} Rg2
{-8.99/11 0.18} 60. Qb1+ {+6.40/16 0.42} Kf2 {-8.82/10 0.13} 61. Qb6+
{+6.40/14 0.09} Kxg3 {-8.15/11 0.08} 62. Qe3+ {+6.55/12 0.11} Kh2+
{-7.31/5 0.00} 63. Kh4 {+6.61/14 0.07} Rb2 {-9.97/13 0.19} 64. Qg3+
{+6.70/15 0.12} Kh1 {-7.52/5 0.00} 65. Qd3 {+7.24/13 0.29} Rb8 {-10.57/14 0.16}
66. Qd5+ {+M15/21 0.14} Kg1 {-12.76/15 0.17} 67. Kg3 {+M11/27 0.15} Rg8+
{-250.00/10 0.01} 68. Qxg8 {+M9/38 0.09} Kf1 {-250.00/14 0.02} 69. Kf3
{+M7/93 0.13} Ke1 {-250.00/13 0.01} 70. Qd5 {+M5/127 0.03} h4
{-M4/16 0.01 White wins by adjudication} 1-0
[/pgn]

PlentyChess 8.0.0 - Komodo Dragon 3.3 (round 4, 1-0): identical to the real game up to 40.Qa8, then 40...f5 and the fastest win among the games that reached move 40 (adjudicated at move 53).

[pgn]
[Event "Kasparov-Topalov"]
[Site "Intel(R) Core(TM) i5-4570 3.20GHz 8GB RAM"]
[Date "2026.11.01"]
[Round "4"]
[White "PlentyChess 8.0.0 x64(BMI2-5Pct)"]
[Black "Komodo Dragon 3.3 x64(AVX2-5Pct)"]
[Result "1-0"]
[BlackElo "2771"]
[ECO "B07"]
[Opening "Pirc defense"]
[Time "01:19:10"]
[Duration "00:00:14"]
[WhiteElo "2844"]
[TimeControl "7+0.1"]
[Termination "adjudication"]
[PlyCount "106"]

1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 Bg7 5. Qd2 c6 6. f3 b5 7. Nge2 Nbd7 8. Bh6
Bxh6 9. Qxh6 Bb7 10. a3 e5 11. O-O-O Qe7 12. Kb1 a6 13. Nc1 O-O-O 14. Nb3 exd4
15. Rxd4 c5 16. Rd1 Nb6 17. g3 Kb8 18. Na5 Ba8 19. Bh3 d5 20. Qf4+ Ka7 21. Rhe1
d4 22. Nd5 Nbxd5 23. exd5 Qd6 24. Rxd4 {-0.31/18 0.53} cxd4 {+2.01/15 0.25} 25.
Re7+ {+0.52/17 0.38} Kb6 {+2.11/18 0.17} 26. Qxd4+ {+0.59/20 0.23} Kxa5
{+2.28/19 0.28} 27. b4+ {+0.54/20 0.27} Ka4 {-0.54/17 0.24} 28. Qc3
{+0.78/22 0.31} Qxd5 {-0.42/20 0.26} 29. Ra7 {+0.45/21 0.49} Bb7
{-0.42/19 0.23} 30. Rxb7 {+0.66/20 0.20} Qc4 {-0.37/17 0.22} 31. Qxf6
{+0.39/17 0.30} Kxa3 {-0.36/19 0.50} 32. Qxa6+ {+3.16/17 0.26} Kxb4
{-0.36/5 0.00} 33. c3+ {+3.08/17 0.25} Kxc3 {-1.54/15 0.16} 34. Qa1+
{+3.44/18 0.22} Kd2 {-3.19/17 0.23} 35. Qb2+ {+3.84/18 0.24} Kd1
{-3.37/17 0.25} 36. Bf1 {+3.71/17 0.37} Rd2 {-3.55/16 0.18} 37. Rd7
{+4.36/17 0.53} Rxd7 {-3.55/15 0.12} 38. Bxc4 {+4.24/16 0.18} bxc4
{-3.79/15 0.13} 39. Qxh8 {+4.47/16 0.30} Rd3 {-3.30/15 0.13} 40. Qa8
{+4.86/14 0.33} f5 {-3.94/15 0.44} 41. Qa4+ {+5.09/14 0.27} Ke1 {-3.56/16 0.34}
42. Qxc4 {+5.62/15 0.30} Rxf3 {-4.17/14 0.10} 43. Qh4 {+6.06/12 0.24} Kf1
{-4.11/13 0.12} 44. Qxh7 {+6.08/14 0.32} g5 {-5.41/16 0.21} 45. h4
{+6.30/12 0.32} g4 {-5.94/15 0.18} 46. Qd7 {+6.36/12 0.25} Kg1 {-4.69/15 0.29}
47. h5 {+6.84/13 0.26} f4 {-8.00/14 0.14} 48. gxf4 {+7.03/12 0.17} g3
{-9.18/14 0.44} 49. h6 {+7.29/13 0.15} Rf1+ {-11.13/13 0.20} 50. Kc2
{+7.39/15 0.20} Rf2+ {-12.67/12 0.10} 51. Kd3 {+7.69/14 0.13} Rf1
{-15.82/15 0.43} 52. Qg4 {+8.01/16 0.28} Rb1 {-250.00/17 0.11} 53. Qxg3+
{+989.00/23 0.10} Kf1 {-250.00/17 0.02 White wins by adjudication} 1-0
[/pgn]

Round-Robin Standings — 30 Engines, 4 Rounds

Code: Select all

+------+-----------------------------------+--------+-------+----------+-----------+-------+--------+--------+
| Rank | Name                              | Points | Games | Live Elo | Start Elo |   +/- | Score% |    S-B |
+------+-----------------------------------+--------+-------+----------+-----------+-------+--------+--------+
|    1 | Reckless 0.9.0 x64(AVX2-5Pct)     |   74,0 |   116 |   2847,3 |    2847,0 |  +0,3 |   63,8 | 3984,0 |
|    2 | PlentyChess 8.0.0 x64(BMI2-5Pct)  |   73,5 |   116 |   2844,2 |    2844,0 |  +0,2 |   63,4 | 3979,2 |
|    3 | Obsidian 16.0 x64(BMI2-5Pct)      |   72,0 |   116 |   2800,3 |    2798,0 |  +2,3 |   62,1 | 3927,5 |
|    4 | Cinder 0.6.1 x64(AVX2-5Pct)       |   71,5 |   116 |   2792,5 |    2790,0 |  +2,5 |   61,6 | 3876,8 |
|    5 | Berserk 14.0 x64(AVX2-5Pct)       |   71,5 |   116 |   2807,6 |    2806,0 |  +1,6 |   61,6 | 3865,0 |
|    6 | Viridithas 20.0.0 x64(AVX2-5Pct)  |   71,0 |   116 |   2817,9 |    2817,0 |  +0,9 |   61,2 | 3829,2 |
|    7 | Clover 9.1 x64(AVX2-5Pct)         |   70,0 |   116 |   2771,3 |    2768,0 |  +3,3 |   60,3 | 3815,8 |
|    8 | Horsie 1.1 x64(AVX2-5Pct)         |   70,0 |   116 |   2753,4 |    2749,0 |  +4,4 |   60,3 | 3780,0 |
|    9 | Stormphrax 8.0.0 x64(BMI2-5Pct)   |   69,5 |   116 |   2777,7 |    2775,0 |  +2,7 |   59,9 | 3763,8 |
|   10 | Tarnished 5.0 x64(AVX2-5Pct)      |   69,5 |   116 |   2746,6 |    2742,0 |  +4,6 |   59,9 | 3761,5 |
|   11 | Halogen 16.0 x64(BMI2-5Pct)       |   69,0 |   116 |   2724,7 |    2719,0 |  +5,7 |   59,5 | 3756,5 |
|   12 | Alexandria 8.1.12 x64(BMI2-5Pct)  |   69,0 |   116 |   2820,0 |    2820,0 |  +0,0 |   59,5 | 3754,8 |
|   13 | Igel 3.8.0 x64(BMI2-5Pct)         |   69,0 |   116 |   2746,4 |    2742,0 |  +4,4 |   59,5 | 3738,5 |
|   14 | Minke 7.0.0 x64(BMI2-5Pct)        |   67,0 |   116 |   2753,3 |    2750,0 |  +3,3 |   57,8 | 3650,0 |
|   15 | Revenge 4.0 x64(BMI2-5Pct)        |   67,0 |   116 |   2704,2 |    2698,0 |  +6,2 |   57,8 | 3646,0 |
|   16 | Velvet 8.1.1 x64(AVX2-5Pct)       |   66,5 |   116 |   2717,3 |    2712,0 |  +5,3 |   57,3 | 3595,2 |
|   17 | Booot 7.4 x64(BMI2-5Pct)          |   62,0 |   116 |   2686,5 |    2681,0 |  +5,5 |   53,4 | 3380,0 |
|   18 | RubiChess 20240817 x64(BMI2-5Pct) |   60,0 |   116 |   2778,3 |    2779,0 |  -0,7 |   51,7 | 3131,5 |
|   19 | Caissa 2.0.0 x64(BMI2-5Pct)       |   52,0 |   116 |   2814,3 |    2820,0 |  -5,7 |   44,8 | 2869,2 |
|   20 | Rebel 16.3 x64(5Pct)              |   50,0 |   116 |   2662,6 |    2660,0 |  +2,6 |   43,1 | 2794,5 |
|   21 | Tcheran 13.0 x64(AVX2-5Pct)       |   49,5 |   116 |   2728,5 |    2730,0 |  -1,5 |   42,7 | 2627,5 |
|   22 | Koivisto 9.0 x64(AVX2-5Pct)       |   48,0 |   116 |   2712,0 |    2713,0 |  -1,0 |   41,4 | 2643,0 |
|   23 | Devre 7.0 x64(AVX2-5Pct)          |   47,5 |   116 |   2775,0 |    2780,0 |  -5,0 |   40,9 | 2577,2 |
|   24 | Integral 8.0 x64(BMI2-5Pct)       |   46,0 |   116 |   2776,4 |    2782,0 |  -5,6 |   39,7 | 2421,2 |
|   25 | Komodo Dragon 3.3 x64(AVX2-5Pct)  |   44,0 |   116 |   2765,4 |    2771,0 |  -5,6 |   37,9 | 2270,2 |
|   26 | Starzix 6.0 x64(AVX2-5Pct)        |   43,5 |   116 |   2752,0 |    2757,0 |  -5,0 |   37,5 | 2410,2 |
|   27 | Motor 0.9.0 x64(AVX2-5Pct)        |   43,5 |   116 |   2751,1 |    2756,0 |  -4,9 |   37,5 | 2340,8 |
|   28 | Patricia 5.1.0 x64(AVX2-5Pct)     |   34,0 |   116 |   2645,8 |    2648,0 |  -2,2 |   29,3 | 1914,0 |
|   29 | RofChade 3.1 x64(AVX2-5Pct)       |   20,5 |   116 |   2673,5 |    2682,0 |  -8,5 |   17,7 |  913,2 |
|   30 | Clarity 8.0.0 x64(BMI2-5Pct)      |   19,5 |   116 |   2695,8 |    2706,0 | -10,2 |   16,8 |  900,5 |
+------+-----------------------------------+--------+-------+----------+-----------+-------+--------+--------+
Reckless takes the lead by half a point. The order follows move 24 more than it follows Elo: Revenge (2698) and Halogen (2719) are in the top half, while Caissa (2820), Devre (2780) and Integral (2782) are in the bottom half.

Komodo Dragon 3.3[/b] is one of the strongest engines in my list, yet in 4 rounds it has never played 24.Rxd4 with White, and with Black it took the rook every single time.
Top
goni-K26
Posts: 312
Joined: Mon Mar 09, 2026 4:16 pm
Full name: Agron Kovaci

Kasparov - Topalov, Wijk aan Zee 1999(Rounds 5-6)

Post by goni-K26 »

Kasparov - Topalov test: rounds 5-6

Rounds 5 and 6 (857 games) gave the same picture at move 24 as before:

Code: Select all

 White's 24th / Black's reply   Games    1-0    1/2    0-1   White score
 ----------------------------   -----   ----   ----   ----   -----------
 24.Nc6+                          279      7     71    201           15%
 24.Rxd4 Kb6 (declined)           452      2    421     29           47%
 24.Rxd4 cxd4 (accepted)          126     72     49      5           77%

Code: Select all

 Original move      Played instead    Games   White score
 ----------------   ---------------   -----   -----------
 25. Re7+           25. Qxd4+             4            0%
 27. b4+            27. Qc3+              1            0%
 28. Qc3            28. Ra7              15           97%
 30... Qc4          30... Rd6            21           98%
 30... Qc4          30... Rhe8           16           91%
 31... Kxa3         31... Rd1+           16           56%
 33. c3+            33. Bd7              33           55%
 35... Kd1          35... Ke1             7          100%
 39... Rd3          39... Kd2             2          100%
 39... Rd3          39... f5              3          100%
 40. Qa8            40. Qa1               1          100%
 40. Qa8            40. Qc8               2          100%
 40... c3           40... Rb3+            1          100%
 40... c3           40... f5              4          100%
Round 6: Reckless - Integral reaches Topalov's resignation

This is the game I was hoping for. Move by move it follows Kasparov-Topalov up to 39...Rd3. Then Reckless plays its usual 40.Qc8 instead of 40.Qa8, so a simple move-by-move comparison stops at 78 plies. But look at what comes next:

Code: Select all

 Kasparov - Topalov            Reckless - Integral
 40.Qa8   c3                   40.Qc8   c3
 41.Qa4+  Ke1                  41.f4    f5
 42.f4    f5                   42.Qc4   Kd2
 43.Kc1   Rd2                  43.Qa2+  Ke1
 44.Qa7   (Topalov resigned)   44.Qa7   Rd2
                               45.Kc1
After 45.Kc1 the position on the board is exactly the one in which Topalov resigned:

Code: Select all

 White: Kc1, Qa7, pawns f4, g3, h2
 Black: Ke1, Rd2, pawns c3, f5, g6, h7
 Black to move

 FEN: 8/Q6p/6p1/5p2/5P2/2p3P1/3r3P/2K1k3 b - - 0 1
So Integral played all of Topalov's ending moves, ...c3, ...f5, ...Ke1, ...Rd2, in a different order (its king went to e1 via d2), and the game reached the final position one move later (Pos 87). Integral played on, and lost 21 moves later. Topalov knew better.

Other transpositions
  • Stormphrax - Komodo Dragon (round 5) and Clover - Komodo Dragon (round 6): the 35...Ke1 route again, back to the original at Pos 78.
  • Revenge - RubiChess (round 5): 28.Ra7 Bb7 29.Rxb7 Qxd5 30.Qc3 gives the same position as Kasparov's 28.Qc3 Qxd5 29.Ra7 Bb7 30.Rxb7. The game came back to the main line and stayed there until 33.Bd7 (Pos 64).
Also notice Integral's defence 31...Rd1+ (16 games in these rounds). It saves a draw in more than half of the games. Topalov's 31...Kxa3 loses.

[pgn]
[Event "Kasparov-Topalov"]
[Site "Intel(R) Core(TM) i5-4570 3.20GHz 8GB RAM"]
[Date "2026.11.01"]
[Round "6"]
[White "Reckless 0.9.0 x64(AVX2-5Pct)"]
[Black "Integral 8.0 x64(BMI2-5Pct)"]
[Result "1-0"]
[BlackElo "2782"]
[ECO "B07"]
[Opening "Pirc defense"]
[Time "08:59:56"]
[Duration "00:00:22"]
[WhiteElo "2847"]
[TimeControl "7+0.1"]
[Termination "adjudication"]
[PlyCount "132"]

1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. Be3 Bg7 5. Qd2 c6 6. f3 b5 7. Nge2 Nbd7 8. Bh6
Bxh6 9. Qxh6 Bb7 10. a3 e5 11. O-O-O Qe7 12. Kb1 a6 13. Nc1 O-O-O 14. Nb3 exd4
15. Rxd4 c5 16. Rd1 Nb6 17. g3 Kb8 18. Na5 Ba8 19. Bh3 d5 20. Qf4+ Ka7 21. Rhe1
d4 22. Nd5 Nbxd5 23. exd5 Qd6 24. Rxd4 {-0.37/20 1.80} cxd4 {+1.59/17 0.81} 25.
Re7+ {+1.17/15 0.24} Kb6 {+2.05/15 0.42} 26. Qxd4+ {+1.43/16 0.23} Kxa5
{+2.08/15 0.25} 27. b4+ {+1.48/16 0.25} Ka4 {+2.27/13 0.15} 28. Qc3
{+1.72/20 0.88} Qxd5 {-0.26/22 0.40} 29. Ra7 {+1.78/17 0.21} Bb7
{-0.30/19 0.14} 30. Rxb7 {+1.23/16 0.24} Qc4 {-0.32/20 0.19} 31. Qxf6
{+1.40/16 0.15} Kxa3 {-0.31/20 0.22} 32. Qxa6+ {+3.72/16 0.17} Kxb4
{-0.28/20 0.20} 33. c3+ {+3.60/14 0.23} Kxc3 {-3.69/19 0.43} 34. Qa1+
{+3.68/11 0.13} Kd2 {-3.30/19 0.14} 35. Qb2+ {+3.76/12 0.22} Kd1
{-3.39/18 0.22} 36. Bf1 {+3.94/13 0.13} Rd2 {-3.96/19 0.34} 37. Rd7
{+3.92/14 0.24} Rxd7 {-3.45/15 0.12} 38. Bxc4 {+4.54/14 0.40} bxc4
{-4.68/19 0.75} 39. Qxh8 {+4.86/13 0.14} Rd3 {-4.44/15 0.32} 40. Qc8
{+4.89/12 0.18} c3 {-4.00/16 0.27} 41. f4 {+5.65/11 0.15} f5 {-6.13/19 1.30}
42. Qc4 {+5.73/12 0.17} Kd2 {-6.78/20 1.40} 43. Qa2+ {+5.98/11 0.12} Ke1
{-6.69/14 0.13} 44. Qa7 {+5.99/13 0.20} Rd2 {-7.17/18 0.74} 45. Kc1
{+6.06/13 0.14} Ke2 {-7.18/16 0.09} 46. Qxh7 {+6.12/17 0.38} Kd3
{-7.57/14 0.12} 47. Qd7+ {+6.11/16 0.20} Kc4 {-7.15/14 0.03} 48. Qf7+
{+6.17/15 0.19} Kb4 {-7.55/16 0.14} 49. h4 {+6.26/14 0.13} Rh2 {-6.43/16 0.07}
50. Qd5 {+6.16/15 0.16} Rd2 {-7.52/15 0.20} 51. Qe6 {+6.15/15 0.15} Rh2
{-7.69/13 0.05} 52. Qc6 {+6.25/13 0.09} Rd2 {-8.12/13 0.15} 53. Qxg6
{+6.11/16 0.21} Rh2 {-7.98/14 0.10} 54. Qc6 {+6.29/14 0.10} Rd2 {-8.08/12 0.08}
55. h5 {+6.49/14 0.13} Rh2 {-8.09/12 0.05} 56. h6 {+6.49/14 0.20} Ka5
{-8.40/15 0.15} 57. Qc5+ {+6.61/16 0.12} Ka6 {-8.15/12 0.03} 58. Qc6+
{+8.63/16 0.62} Ka7 {-8.15/13 0.05} 59. Qd7+ {+9.39/9 0.11} Kb6 {-8.64/14 0.23}
60. h7 {+10.11/10 0.12} Rh1+ {-8.08/11 0.03} 61. Kc2 {+10.38/12 0.07} Rh2+
{-9.36/14 0.11} 62. Kxc3 {+199.87/14 0.12} Rh6 {-9.39/14 0.11} 63. Qd4+
{+199.91/16 0.20} Kc6 {-11.98/14 0.16} 64. h8=Q {+199.93/17 0.71} Rxh8
{-325.03/23 0.06} 65. Qxh8 {+199.95/15 0.13} Kd6 {-325.05/21 0.03} 66. Qe5+
{+199.97/15 0.24} Kc6 {-325.07/19 0.03 White wins by adjudication} 1-0
[/pgn]

Standings after round 6

PlentyChess is back on top, half a point ahead of Reckless. The order hardly moved in these two rounds.

Round-Robin Standings — 30 Engines, 6 Rounds

Code: Select all

+------+-----------------------------------+--------+-------+----------+-----------+-------+--------+--------+
| Rank | Name                              | Points | Games | Live Elo | Start Elo |   +/- | Score% |    S-B |
+------+-----------------------------------+--------+-------+----------+-----------+-------+--------+--------+
|    1 | PlentyChess 8.0.0 x64(BMI2-5Pct)  |  111,5 |   174 |   2844,8 |    2844,0 |  +0,8 |   64,1 | 9028,5 |
|    2 | Reckless 0.9.0 x64(AVX2-5Pct)     |  111,0 |   174 |   2847,4 |    2847,0 |  +0,4 |   63,8 | 8931,5 |
|    3 | Cinder 0.6.1 x64(AVX2-5Pct)       |  109,0 |   174 |   2794,4 |    2790,0 |  +4,4 |   62,6 | 8867,8 |
|    4 | Berserk 14.0 x64(AVX2-5Pct)       |  108,5 |   174 |   2808,9 |    2806,0 |  +2,9 |   62,4 | 8826,8 |
|    5 | Obsidian 16.0 x64(BMI2-5Pct)      |  108,0 |   174 |   2801,4 |    2798,0 |  +3,4 |   62,1 | 8840,2 |
|    6 | Viridithas 20.0.0 x64(AVX2-5Pct)  |  107,0 |   174 |   2818,5 |    2817,0 |  +1,5 |   61,5 | 8635,2 |
|    7 | Alexandria 8.1.12 x64(BMI2-5Pct)  |  105,0 |   174 |   2820,5 |    2820,0 |  +0,5 |   60,3 | 8558,0 |
|    8 | Horsie 1.1 x64(AVX2-5Pct)         |  105,0 |   174 |   2755,5 |    2749,0 |  +6,5 |   60,3 | 8455,2 |
|    9 | Stormphrax 8.0.0 x64(BMI2-5Pct)   |  104,5 |   174 |   2779,1 |    2775,0 |  +4,1 |   60,1 | 8517,0 |
|   10 | Tarnished 5.0 x64(AVX2-5Pct)      |  104,5 |   174 |   2748,9 |    2742,0 |  +6,9 |   60,1 | 8457,5 |
|   11 | Clover 9.1 x64(AVX2-5Pct)         |  104,0 |   174 |   2772,5 |    2768,0 |  +4,5 |   59,8 | 8478,0 |
|   12 | Minke 7.0.0 x64(BMI2-5Pct)        |  102,5 |   174 |   2755,6 |    2750,0 |  +5,6 |   58,9 | 8339,0 |
|   13 | Igel 3.8.0 x64(BMI2-5Pct)         |  101,5 |   174 |   2747,9 |    2742,0 |  +5,9 |   58,3 | 8249,8 |
|   14 | Halogen 16.0 x64(BMI2-5Pct)       |  101,0 |   174 |   2726,7 |    2719,0 |  +7,7 |   58,0 | 8274,2 |
|   15 | Revenge 4.0 x64(BMI2-5Pct)        |  100,0 |   174 |   2707,1 |    2698,0 |  +9,1 |   57,5 | 8189,5 |
|   16 | Velvet 8.1.1 x64(AVX2-5Pct)       |   99,5 |   174 |   2719,7 |    2712,0 |  +7,7 |   57,2 | 8053,8 |
|   17 | Booot 7.4 x64(BMI2-5Pct)          |   94,0 |   174 |   2689,5 |    2681,0 |  +8,5 |   54,0 | 7642,2 |
|   18 | RubiChess 20240817 x64(BMI2-5Pct) |   89,0 |   174 |   2777,6 |    2779,0 |  -1,4 |   51,1 | 6934,5 |
|   19 | Caissa 2.0.0 x64(BMI2-5Pct)       |   79,0 |   174 |   2811,9 |    2820,0 |  -8,1 |   45,4 | 6500,0 |
|   20 | Tcheran 13.0 x64(AVX2-5Pct)       |   76,0 |   174 |   2728,4 |    2730,0 |  -1,6 |   43,7 | 6089,0 |
|   21 | Rebel 16.3 x64(5Pct)              |   74,5 |   174 |   2663,7 |    2660,0 |  +3,7 |   42,8 | 6247,2 |
|   22 | Devre 7.0 x64(AVX2-5Pct)          |   73,0 |   174 |   2773,2 |    2780,0 |  -6,8 |   42,0 | 5934,2 |
|   23 | Koivisto 9.0 x64(AVX2-5Pct)       |   69,5 |   174 |   2710,7 |    2713,0 |  -2,3 |   39,9 | 5675,5 |
|   24 | Motor 0.9.0 x64(AVX2-5Pct)        |   69,0 |   174 |   2749,9 |    2756,0 |  -6,1 |   39,7 | 5524,5 |
|   25 | Integral 8.0 x64(BMI2-5Pct)       |   66,5 |   174 |   2772,9 |    2782,0 |  -9,1 |   38,2 | 5267,0 |
|   26 | Komodo Dragon 3.3 x64(AVX2-5Pct)  |   64,5 |   174 |   2762,2 |    2771,0 |  -8,8 |   37,1 | 4932,0 |
|   27 | Starzix 6.0 x64(AVX2-5Pct)        |   64,0 |   174 |   2749,2 |    2757,0 |  -7,8 |   36,8 | 5266,0 |
|   28 | Patricia 5.1.0 x64(AVX2-5Pct)     |   49,0 |   174 |   2644,1 |    2648,0 |  -3,9 |   28,2 | 4136,0 |
|   29 | RofChade 3.1 x64(AVX2-5Pct)       |   32,5 |   174 |   2669,9 |    2682,0 | -12,1 |   18,7 | 2204,0 |
|   30 | Clarity 8.0.0 x64(BMI2-5Pct)      |   27,0 |   174 |   2690,1 |    2706,0 | -15,9 |   15,5 | 1840,8 |
+------+-----------------------------------+--------+-------+----------+-----------+-------+--------+--------+
Top

Return to “Computer Chess Club: Tournaments and Matches”