1997 was the year when the machines took over. It was when the super computer Deep Blue defeated Gary Kasparov in a match over 6 games by 3,5 to 2,5. The first defeat of a reigning world chess champion to a computer under tournament conditions. However, it was marginal and Kasparov might still have been better than his robot opponent at that time. But chances are, that before the New Millennium began, computers were definitively better than any man under the sun.
At the CEBIT tournament in Hannover in 1991, I had the chance to play against the direct predecessor of Deep Blue, which was running under the name of Deep Thought. IBM sponsored the event, maybe hoping that their brainchild would already be in a state, to outclass the participating German Grandmasters. But this didn’t happen. The tournament was won by the author and Deep Thought only finished 7th in a field of 8 players. The machines still needed to wait a couple of years more, before their time would finally come……
[Event "Hannover"]
[Site "Hannover"]
[Date "1991.??.??"]
[Round "6"]
[White "Comp Deep Thought"]
[Black "Wahls, Matthias"]
[Result "0-1"]
[ECO "E86"]
[BlackElo "2560"]
[PlyCount "56"]
[EventDate "1991.03.??"]
[EventType "tourn"]
[EventRounds "7"]
[EventCountry "GER"]
[EventCategory "9"]
[SourceTitle "CBM 22"]
[Source "ChessBase"]
1. d4 d6 2. c4 g6 3. Nc3 Bg7 4.
e4 Nf6 5. f3 {Diagram [#] Now the opening is defined: King's Indian Saemisch
Variation.} O-O 6. Be3 e5 7. Nge2 c6 8. Qd2 Nbd7 9. d5 (9. O-O-O {is the other
mainline here.}) 9... cxd5 10. Nxd5 {Diagram [#] Deep Thought opts for a
sideline. Most probably, he was fed with this specific variation, because the
arising pawn structure is more open compared to the closed centre resulting
from 10.cxd5.} Nxd5 11. Qxd5 ({It makes absolutely no sense to recapture with
the pawn. After} 11. cxd5 f5 {Black was able to mobilize his f-pawn quicker
compared to the mainline 10.cxd5. Also, the exchange of the minor pieces tends
to help the side with the lesser space.}) 11... Nb6 {Diagram [#] During the game, I
was not happy at all with the result of the opening. First of all, I was out
my books after 10.Nxd5 (Note that I was no experienced King's Indian player. I
started my liaison with the King's Indian only two years before this game).
Secondly, my game plan was to build up a closed position, where maneuvering
and the application of positional principles would be on the menu. Now however, the position was open to some degree and the calculative power of the machine had become an important factor.} 12. Qb5 {A common match-up has arisen:
One side has a static advantage, the other side a dynamic one. If White would
be able to finish his development and bring his king into safety, his superior
pawn structure would start to tell. Black has to exploit his lead in development in
order to prevent this from happening. He will either try to attack the king or
force a favorable change of structure. For this reason my only choice was to
play actively in a very determined fashion.} Bh6 $1 {Diagram [#] An important
strategic device, which improves Black's position, no matter what White will
do. If he exchanges on h6, the weakness of the dark color complex will make
itself felt. If he retreats the bishop to f2, Black will have activated his own
bishop with a gain of tempo and prevented White from castling queenside.} 13.
Bf2 (13. Bxh6 $6 Qh4+ 14. Ng3 Qxh6 {[%cal Gh6e3]}) 13... Be6 14. Nc3 Qc7 $1 {
The last three moves were very energetic and all contained threats. That's the
way how you want to develop your pieces. Now White has to weaken the square c3,
which will play an important role later in the game.} 15. b3 Nd7 {Diagram [#]
This was a novelty at the time. The idea behind it is logical. In order to
exploit my lead in development, I have to open up the position by applying a
pawn lever. Since 15..f5? 16.Bd3 doesn't fit the bill, it has to be b7-b5.
Well, as the engines indicate today, there even is a third pawn thrust
conceivable:} (15... d5 $5 16. exd5 Nxd5 17. Nxd5 Bxd5 18. Qxd5 Rad8 {But now,
there are lines where White can force a draw by repetition, so I wouldn't call
it an improvement.}) 16. Qb4 a6 17. Rd1 Rfc8 {It was possible to play b5
outright:} (17... b5 18. Nd5 (18. Qxd6 Qa5 19. b4 Qa3 $19) (18. cxb5 $2 a5 $19)
18... Bxd5 19. Rxd5 Rfc8 {with transposition to the game, or alternatively} (
19... Nc5 20. Qa3 Ne6 21. g3 Nd4 22. Bxd4 exd4 23. Bd3 Be3 24. Kf1 $13)) 18.
Nd5 (18. Qxd6 $2 Qa5 19. Qd3 (19. b4 Qa3 $19) 19... b5 $19) (18. a4 Bf8 19. Nd5
Bxd5 20. exd5 (20. Rxd5 $2 Nf6 21. Rd1 d5 $17) 20... Nc5 21. Bxc5 dxc5 22. Qd2
Bd6 23. g3 f5 {with an ongoing initiative}) 18... Bxd5 19. Rxd5 b5 ({An
interesting alternative was} 19... Bf8 $5 {with the idea to exploit the
location of White's queen by following up with 20..Nf6 and 21..d5.} 20. Qa5 (
20. Qc3 $2 Nf6 21. Rd1 d5 $1 22. exd5 a5 $17) 20... Qxa5+ 21. Rxa5 b5 {Black
has the initiative, but as it appears, White can hold his head above the water
in every line:} 22. Kd1 (22. Be2 $2 d5 $19) (22. b4 $5 bxc4 (22... d5 23. c5
dxe4 24. fxe4 Nxc5 25. bxc5 Rxc5 $1 {The bishop is still needed to pick up the
stranded rook.} 26. Bxc5 Bxc5 27. Bxb5 $1 {The only move.} Bb4+ 28. Ke2 Bxa5
29. Bc4 {and White should manage to safe the draw.}) 23. Kd2 d5 (23... Rab8 24.
a3 Rc6 25. Kc3 Be7 26. Be2 Bd8 27. Rd5 Nb6 28. Bxb6 Bxb6 29. Rb1 Bd4+ 30. Kc2
Kg7 31. a4) 24. Rxd5 Bxb4+ 25. Kc2 Bc5 26. Rxd7 Bxf2 27. Be2 Rab8 28. Rb1 $15)
22... d5 (22... Nc5 $11) 23. cxb5 axb5 (23... Bh6 24. Rxa6 Rc1+ 25. Ke2 Rac8
26. Rc6 Rc2+ 27. Ke1 R8xc6 28. bxc6 Rxc6 29. exd5 Rc1+ 30. Ke2 Rc2+ 31. Ke1
Rxa2 32. Rg1 Ra1+ 33. Ke2 Nf6 34. Kd3 Rd1+ (34... Nxd5 35. Ke4 $11) 35. Kc4 Rd2
36. Ba7 Nxd5 37. Kb5 $13) 24. Rxa8 Rxa8 25. Bxb5 Rxa2 26. Re1 Rxf2 (26... dxe4
27. Re2 Ra1+ 28. Kc2 Nf6 29. Kb2 exf3 30. gxf3 Rf1 31. Rc2 $13) 27. Bxd7 Rxg2
28. Re2 Rg1+ 29. Re1 Rxe1+ 30. Kxe1 dxe4 31. fxe4 Kg7 32. Ke2 Kf6 33. Kf3 Kg5
34. Bc8 Bc5 35. Bd7 f5 36. Be6 {with a probable draw}) 20. a4 $2 {Diagram [#]
This is a very instructive mistake, which highlights one of the general
problems of computers, even in present times. Deep Thought calculated that the immediate 20.Qxd6 fails to 20..Qa5 21.b4 Qxa2 -+. It intended to
improve on that by inserting 20.a4 bxa4, thus blocking the path to a2. What it
didn't see, however, was the following piece sacrifice. It couldn't do so, because it was a long term sacrifice geared to exploit the awkward location of White's
king. In order to appreciate this, the machine would have had to calculate
every line until its end, which would be mate in some occasions. Due to the
exponential explosion of variation trees in complicated situations, however,
this was a task too big to manage. The respective mating positions simply lay
behind its horizon of calculation.} (20. Qxd6 $2 Qa5+ 21. b4 Qxa2 22.
Qxd7 Qb1+ 23. Rd1 Qxb4+ 24. Ke2 Qxc4+ 25. Qd3 Qa2+ 26. Ke1 Qa5+ 27. Ke2 Rd8 $19
) (20. cxb5 $2 Qc1+ 21. Rd1 a5 22. Qxd6 Qc3+ 23. Ke2 Qb2+ 24. Ke1 Bf8 25. Qxd7
Bb4+ 26. Rd2 Rd8 $19) (20. Rd1 $6 Nc5 21. Qc3 Rcb8 22. Be2 b4 23. Qc2 a5 24.
O-O a4 25. Rb1 axb3 26. axb3 Ra3 $15) (20. Be2 Bf8 (20... Nc5 21. O-O a5 22.
Qe1 $1 b4 23. Bd1 $1 {and White can prevent a4}) 21. cxb5 Nf6 22. Rd1 d5 23.
Qd2 axb5 24. Bxb5 Qb7 (24... dxe4 25. O-O exf3 26. gxf3 e4 27. a4 exf3 28. Kh1)
(24... Qc2 25. Qxc2 Rxc2 26. a4 Bb4+ 27. Kf1 dxe4 28. h4 $13) 25. a4 dxe4 26.
O-O exf3 27. gxf3 Qxf3 28. Qd3 Qxd3 29. Rxd3 Ne4 $15) 20... bxa4 21. Qxd6 (21.
Qxa4 Nb6 22. Bxb6 Qxb6 $19) 21... Qb7 $1 {Diagram [#] After convincing myself,
that there was no victory based on my passed pawn after the exchange of queens,
I focused on mating my opponent.} 22. bxa4 (22. Qxd7 Qb4+ 23. Ke2 axb3 $19) (
22. Be2 Rd8 23. Qa3 Qxb3 24. Rd3 Qc2 25. Qc3 Qxc3+ 26. Rxc3 Nc5 27. O-O Bf8 $19
) 22... Bf8 $1 23. Qxd7 Qb4+ $1 24. Rd2 {Maybe a human would have gone for 24.
Kd1 in order to pose some last practical problems. The machine, however,
discarded it as being inferior by objective standards and settled for a
completely hopeless position. Admittedly, the task after 24.Kd1 wouldn't have
been too challenging:} (24. Kd1 Qb3+ 25. Kd2 (25. Ke2 Qxc4+ 26. Ke3 Qc1+ 27.
Ke2 Bb4 $19 {[%cal Rc8c2]}) 25... Bb4+ 26. Kc1 Rcb8 $1 $19) 24... Rd8 25. Qxd8
Rxd8 26. Be3 Bc5 27. Bg5 Rd6 28. Ke2 Rxd2+ {Diagram [#] ...and while I was
pondering, whether I should leave the c-pawn alive after 29.Bxd2, in
order not to activate his light-squared bishop, or better take both of the
pawns, the representative of IBM came to my board and congratulated me for my
victory. What I felt delighted about was the fact, that I didn't beat the
machine as a result of a long-winded positional struggle, which was my game
plan, but on his very own territory, instead.} 0-1