Author: Stephen A. Boak
Date: 01:39:06 09/26/00
Go up one level in this thread
Here's the promised look at the Deep Thought - Fishbein game, where Fishbein
eventually lost the endgame with KR v KRB held by Deep Thought.
Deep Thought - Fishbein,A [C69]
Los Angeles, California, 1988
{Steve Boak move notes, assisted by Fritz6a (plus 5-piece tablebases)}
1.e4 C69: Ruy Lopez: Exchange Variation with 5 0-0 f6 1...e5 2.Nf3 Nc6 3.Bb5 a6
4.Bxc6 dxc6 5.0-0 f6 6.d4 exd4 7.Nxd4 c5 8.Ne2 Qxd1 9.Rxd1 Bd7 10.Bf4 0-0-0
11.c4 Ne7 12.Nbc3 Re8 13.Be3 Nc6 14.Rd2 Ne5 15.b3 h5 16.h3 b6 17.Nd5 Bc6 18.a4
a5 19.Nec3 Kb7 20.Nb5 Bxb5 21.axb5 Bd6 22.f3 Ra8 23.f4 Nf7 24.Kf2 Rhe8 25.Kf3
Bf8 26.Nc3 Nd6 27.Rd5 g6 28.g4 hxg4+ 29.hxg4 Bg7 30.f5 gxf5 31.gxf5 Rh8 32.Bf4
Rh3+ 33.Bg3 Rg8 34.Rd3 Bf8 35.Kg2 Rh4 36.Re1 Rhg4 37.Kh3 a4 [37...Nf7!] 38.bxa4
R8g7?? 1. ± (1.00): 38...R4g5 39.Nd5 Bg7 40.Nf4 Nxc4 41.Ne6 Ne5 42.Rd5 Rh5+
43.Kg2
2. ± (1.09): 38...Nxc4 39.Nd5 Ne5 40.Bxe5 fxe5 41.Nf6 Rg1 42.Rxg1 Rxg1 43.Rd5
39.Bh4 Nxc4? 1. +- (1.53): 39...Rxh4+ 40.Kxh4 Nxc4 41.Nd5 Bd6 42.Kh3 Nb2 43.Ra3
Nc4 44.Rb3
2. +- (1.94): 39...Be7 40.Nd5 Nxc4 41.Bxf6 Bxf6 42.Nxf6 R4g5 43.Rd7 Rxd7 44.Nxd7
40.Bxf6 White has a 2-point edge, per Fritz6a. 40...Rg8 41.e5 Bh6 42.Nd5 Bf4
43.Rf3 Bxe5 44.Rxe5 1. +- (1.78): 44.Rxe5 Nxe5 45.Bxe5 Re4 46.Re3 Rxa4 47.Nxc7
Rg5 48.Rf3 Re4
2. ² (0.28): 44.a5 Nd2 45.a6+ Ka7 46.Rfe3 Bf4 47.Ne7 Rg3+ 48.Kh4 R8g4+
3. = (-0.12): 44.Rd3 Bd6 45.Rf1 c6 46.Ne7 Rg3+ 47.Rxg3 Rxg3+ 48.Kh4 cxb5
4. = (0.00): 44.Bxe5 Nxe5 45.Rxe5 Rg1 46.Kh2 R8g2+ 47.Kh3 Rg8 44...Nxe5 45.Bxe5
R4g5 46.Bg3 c4 47.Nc3 Rh5+ 48.Kg2 Rgg5 49.f6 Rf5 50.Re3 Rxf6 51.Re7 Rff5
52.Rxc7+ Ka8 53.a5 Fritz6a believes there are many much better moves, preserving
approximately a 3-point edge for White. But Deep Thought gives back 2-points,
keeping only a 1-point White edge! Fritz6a examples of better moves:
1. +- (3.03): 53.Rc6 Rhg5
2. +- (2.91): 53.Ne4
3. +- (2.75): 53.Rc8+ Ka7 54.Ne4
4. +- (2.59): 53.Rxc4 Rc5 54.Rxc5 Rxc5 55.Be1 Kb7 56.Kf3 Rc8 57.Bd2 Rd8
5. +- (2.53): 53.Re7 53...Rxb5 54.Nxb5 Rxb5 55.axb6 Rxb6 56.Bf4 Rb3 57.Kf2 Rb7
58.Rxc4 With this capture, Deep Thought enters a tablebase draw. 58...Rb2+
59.Ke3 Kb7 60.Ke4 Rb5 61.Be3 Rh5 62.Bd4 Rg5 63.Be5 Kb6 64.Kd5 Kb5 65.Rc8 Rg6
66.Rc5+ Kb6 67.Rc2 Rh6 68.Rb2+ Ka6 69.Bd6 Rh5+?? Oops! Fishbein enters a
tablebase loss position (Black gets mated in 13 moves!). Only Ka5! keeps a
tablebase draw. Somehow Deep Thought got a position (obviously Fishbein choices
helped lead to this point--but would such a knife-edge juncture have occurred
for the human sooner or later anyway?) where there was only one move for the
human to hold the draw, and Fishbein calculated or guessed wrong. 70.Kc6 Rg5
71.Rb3 Rh5 72.Rb1 non-optimal: mates in 13; Rb1 mates in 11. 72...Rg5 73.Rb2 Rh5
74.Rb8 Rh7 75.Bf8 Ka7 returns the favor with a non-optimal defense: Rf7 lasts
longest (mate in 9), whereas text allows mate in 7. 76.Rb3 Ka8 77.Re3 Rb7 78.Re4
Rf7 79.Re8+ Ka7 80.Bc5+ Ka6 81.Ra8+ mating after 82...Ra7 Rxa7++ 1-0
Bottom line thoughts (calculations from Fritz6a)--Fishbein (an IM at the time)
did very well against the monster computer, Deep Thought. He was struggling in
the opening (down 0.4 to 0.5 pawns at times) and middlegame, but somehow hung in
there with the machine.
On move 37 Fishbein missed Nf7! which would have left him down a mere and
manageable 1/5 or 1/4 pawn. On move 53, Deep Thought strangely played a5?,
giving back about 2/3 of its approx 3-point advantage. Fishbein kept up the
strugggle/complications and obtained an underdog book draw KR v KRB by move 58,
which he held as late as move 68.
Then Fishbein slipped into a mate in 13 lost position, after which Deep Thought
played sub-optimally at one point (losing 2 moves toward shortest mate).
Ironically, Fishbein played sub-optimally once thereafter, giving back Deep
Thought the 2 moves toward mate.
I have heard that Fritz6a doesn't handle tablebases perfectly in some instances.
I don't know if there were any programming glitches in the above situations,
but I think the basic story is sound.
Conclusion (my own): it is not an easy draw for even a strong human to hold
(such as IM Fishbein), especially when the clock is ticking and potential 2nd
time control or sudden death final time control may be looming. Missing a
13-move mate, as Fishbein did, is not unforgiveable for a human, even an IM!
It was enjoyable to compare a strong modern program (Fritz6a) moves with those
of a mega-machine competitor of 11-12 years ago. It would have been nice to
have the clock times available for both Fishbein and Deep Thought, to allow
better comparison with Fritz6a move selections under similar time constraints,
but alas I lack that information.
Thanks again, Ricardo, for posting the game score! Any additional thoughts
after the above analysis?
--Steve
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.