Author: Jeremiah Penery
Date: 11:44:35 04/30/99
Go up one level in this thread
Incedentally, my modified Crafty finds the best move in the first 4 positions
(However, it does not SOLVE the first one) relatively quickly. The rest of the
positions haven't been run yet. Here's some output from my Crafty first, then
the original Nolot analyses (Made by Pierre Nolot, I think), and then some DT2
analysis.
Modified Crafty 16.6
---------------------------------
Position #1:
<This is when it sticks to the move - after 8 minutes, however,
it is not 'solved', because it does not give the correct follow-up or
evaluation>
11 8:03 -1.11 1. Nxh6 c3 2. Re2 Rxh6 3. Nxc3 Qh5
4. Qd2 Nxc3 5. Bxc3 d5 6. Bd4 Qh4
<snip>
15 1076:26 -0.72 1. Nxh6 c3 2. Ba1 Rxh6 3. Nxc3 Qd7
4. Nxd5 Nxd5 5. Rg6 Rxg6 6. Qh5+ Kg8
7. Bxg6 Nf6 8. Bxf6 gxf6 9. Be8 Qd8
Position #2:
<It sticks to the move at about 12 minutes>
11 11:48 ++ 1. Rxc5!!
11 13:56 1.16 1. Rxc5 Nxc5 2. Nf6+ Kh8 3. Qh4 Qb5+
4. Kf2 Nd3+ 5. Kg2 Nf4+ 6. Bxf4 Qe2+
7. Qf2 Qd3 8. Bh6 Rfc8
11-> 14:20 1.16 1. Rxc5 Nxc5 2. Nf6+ Kh8 3. Qh4 Qb5+
4. Kf2 Nd3+ 5. Kg2 Nf4+ 6. Bxf4 Qe2+
7. Qf2 Qd3 8. Bh6 Rfc8
Position #3:
<Sticks in less than 9 seconds - DT2 never found the move>
8 8.62 -0.26 1. Nxg5 Bxd1 2. Nxe6 Qb8 3. Nxg7+ Kd8
4. Ne6+ Kd7 5. Kxd1 Kxe6
8-> 8.89 -0.26 1. Nxg5 Bxd1 2. Nxe6 Qb8 3. Nxg7+ Kd8
4. Ne6+ Kd7 5. Kxd1 Kxe6
<snip>
14 11:32 0.88 1. Nxg5 Bxd1 2. Nxe6 Qb8 3. Nxg7+ Kf7
4. N7f5 Ba4 5. b3 Ke8 6. bxa4 Kd8 7.
Nxe7 Nxe7 8. Bg5
Position #4:
<Finds the move in under 8 seconds>
8 7.59 ++ 1. Nxe6!!
8 8.20 2.61 1. Nxe6 Qxe6 2. Nd5 Qe5 3. f4 Qxb2
4. Nc7+ Ke7 5. Nxa8
<snip>
12 5:54 1.81 1. Nxe6 Qxe6 2. Nd5 Nxd5 3. exd5 Qf5
4. Be3 Nf6 5. Bf3 Ng4 6. Re1 Nxh2 7.
Kxh2
----------------------------------------------------------------------------
Original analysis:
-----------------------------
# Position: 1
# Move: W
r...qb.k Kasparov-Karpov, 20th game 1990
.b....p. 26.Nxh6!! c3 (26... Rxh6!? is not sufficient:
p..pr..p 27.Nxd6 Qh5 (best) 28.Rg5! Qxd1 29.Nf7+ Kg8 30.Nxh6+ Kh8
...n.... 31.Rxd1 c3 32.Nf7+ Kg8 33.Bg6! Nf4 34.Bxc3 Nxg6 35.Bxb4 Kxf7
Pnp.N.N. 36.Rd7+ Kf6 37.Rxg6+ Kxg6 38.Rxb7 +-) 27.Nf5! cxb2
......RP 28.Qg4 Bc8 (if 28... g6!? 29.Kh2! wins : 29...Qd7 30.Nh4 Bc6
.B...PP. 31.Nc5! dxc 32.Rxe6 Nf6 33.Nxg6+ Kg7 34.Qg5 Nbd5 35.Ne5 Kh8
.B.QR.K. 36.Nxd7 +-) 29.Qh4+ Rh6 30.Nxh6 gxh6 31.Kh2! Qe5 32.Ng5 Qf6
33.Re8 Bf5 34.Qxh6 (pretty, but there was a mate in 6 :
34.Nf7+ Qxf7 35.Qxh6+ Bh7 36.Rxa8 Nf6 37.Rxf8 Qxf8 38.Qxf8+ Ng8
39.Qg7#) 34...Qxh6 35.Nf7+ Kh7 36.Bxf5+ Qg6 37.Bxg6+ Kg7 38.Rxa8 Be7
39.Rb8 a5 40.Be4+ Kxf7 41.Bxd5+ 1-0
A very deep combimation, that Tasc R30 or Genius 2 running on a pentium would
take between a few months and a few years to find.
The best Novag computer, the Diablo 68000, finds 26.Nxh6 after 7 months
and a half (Pierre Nolot has let it run on the position for 14 months and one
day, until a power failure stopped an analysis of over 80 000 000 000 nodes!)
but for wrong reasons : it evaluates white's position as inferior and
thinks this move would enable it to draw.
# Position: 2
# Move: W
r....rk. Bronstein - Ljubojevic, Petropolis itz 1973
pp.n.p.p 22.Rxc5!! Nxc5 23.Nf6+ Kh8 24.Qh4 Qb5+ (Now computers have
.nqP..p. a problem here : they think there is perpetual check, but...)
..b.P.B. 25.Ke3! wins 25... h5 26.Nxh5 Qxb3+ and Bronstein won in 41 moves.
....NQ.. If black plays 26... d5+ 27.Bxd5 Qd3 28.Kf2 Ne4+ 29.Bxe4 Qd4+
.B...P.. 30.Kg2 Qxb2+ 31.Kh3 +-
PP..K..P The best programs should be able to find 25.Rxc5!! in a few weeks.
..R.....
# Position: 3
# Move: W
r..qk..r Smaguine - Sahovic, Bienne 1990
ppp.b.pp White wins with a queen sac but black has several ways to defend
..n.p... 12.Nxg5!! Bxd1 13.Nxe6 Qb8 14.Nxg7!! Kf8 15.Bh6! Bg4 16.0-0+
...pP.n. 17.Kg8 17.Rf4 +-
...P..b. It should take between a few months and a few years for a program
..PB.NN. to find 12.Nxg5!!
PP....PP
R.BQK..R
# Position: 4
# Move: W
r.b.kb.r Bronstein - Kotov, candiates, Budapest 1950
.p.n.ppp 10.Nxe6!! Qxe6 11.Nd5 Kd8 12.Bg4 Qe5 13.f4 Qxe4 (13...Qxb2
p..ppn.. was stronger but not sufficient : 14.Bxd7 Bxd7 15.Rb1 Qa3
......BB 16.Nxf6 Bb5 17.Qd4 Qc5 18.Rfd1 +-) 14.Bxd7 Bxd7 15.Nxf6 gxf6
..qNP... 16.Bxf6+ Kc7 17.Bxh8 and Kotov resigned on move 27.
..N..... Tasc R30 (The King), the strongest program for this
PPP..PPP kind of combination, would need a few weeks here.
R..Q.RK.
# Position: 5
# Move: W
r..qrb.k Spassky - Petrossian, w ch (19), Moscow 1969
.p.b..p. 21.e5!! dxe5 22.Ne4! Nh5 23.Qg6!? (This is allright, but
p..ppn.p 23.Qg4!! Nf4 24.Nf3 Qc7 25.Nh4 +- was even stronger) 23...exd4?
........ 24.Ng5 1-0
...NP... Although No program finds 21.e5!! after 8 days, this might well
.BN..... be the first of the 11 positions to be solved by computers in
PPP...QP reasonnable time...
.K...RR.
Well... He sould have said : No program I have tested... as we now know!
# Position: 6
# Move: B
rnbqk..r Malaniouk - Ivantchouk, Moscow 1988
.p...ppp 13... axb5!! offers a rook to keep the white queen out of
p....... play. 14.Qxa8 Bd4 15.Nxd4 cxd4 16.Qxb8 0-0! 17.Ke1 Qh4
.NpPp... 18.g3 Qf6 19.Bf4 g5? (Ivantchouk would have won much faster
QPP.P.n. had he found 19...d3!, a move he found during the analysis.
P....N.. Tasc R30 inds 19...d3! in 2 1/2 hours. But to find 13...axb5
....KbPP it would probably need a couple of centuries...) 20.Rc1 exf4
R.B..B.R 21.Qxf4 Qd4 22.Rd1 bxc4 23.e5 Qc3+ 24.Rd2 Re8 25.Bxd3 cxd3
26.0-0 Nxe5 -+
Last minute info from Pierre, 19...Bf5!! is even stronger than d3
What does DT think about move 19 for black, and in how much time, asks Pierre.*
# Position: 7
# Move: W
.r.bk..r Todorovic - Tosic, Arandjelovac, 1993.
..R..ppp A nice exchange sacrifice to keep control of the 7th rank :
p...p... 1.Rxd8+!! Rxd8 (if 1...Kxd8 2.Ra7! Qe2 3.Qd4+ e8 4.h3 Qe1+
.b..P..q 5.Kh2 Rd8 6.Qc5 Qh4 7.Ba3 Rd7 8.Ra8+ Rd8 9.g3 1-0)
....QP.. 2.Ba3 Qe2 3.h3! Bd7 (better but still losing was 3...Qe1+
....N... 4.Kh2 Qa5 5.Re7+ Kf8 6.Rd7+ Kg8 7.Bb4 Rxd7 8.Bxa5 +-)
.B....PP 4.Nf5! Qd1+ 5.Kh2 f6 6.e5xf6 1-0
...R..K. Genius 2 does better than the R30 on this one, but
needs more than a month to find 1.Rxd8!! even on a 90 MHz
Pentium.
# Position: 8
# Move: W
r...rbk. Goufeld - Osnos, Koutaissi 1978
ppq..ppp After the obvious 24.Bxh7+!! you stil have to find the right
..b.pB.. attacking moves as black has some serious defensive
........ resources. 24...Kxh7 25.Qh5+ Kg8 26.Rd4! gxf6 27.Rg4+ Bg7
......Q. 28.Qh6 Kf8 29.Rxg7! Rac8 (computers prefer 29...Be4 which
.P.B...P also loses after 30.Rg4+ Ke7 31.Rxe4 Rad8 32.c4 Qa5 33.Rae1+-)
P.P..PP. 30.Qh7 b5 31.Rd1 Bd5 32.c4 bxc4 33bxc4 1-0.
R..R..K. It should take the best micros a few months to find 24.Bxh7+!!
As Deep Thought found out, 26 ... gxf6 seems to be a mistake.
More on that later, but this one is dubious now!
# Position: 9
# Move: W
r....r.k Weinstein - Elyoseph, Israel 1992
....bppb This one is really beautiful and should resist computers
..n.p..p for quite a long time, maybe until next century?
p.n.P... 1.Ng5!! hxg5 2.hxg5! Rac8 3.Nf6!! Nb8 (black can also
.p.p.BNP try 3...gxf6, which loses if white play precisely : 4.gxf6
...P.NP. 4...Rfe8 (best defense) 5.Qh5 Rg8 6.Rxc5! Bg6! 7.Qh4 Bxc5
qP..QPB. 8.Be4 Ne7 9Kg2 Qd5 10.Bxd5+-)
..RR..K. 4.Qh5 Bxf6 5.gxf6 gxf6 6.Rxc5 Rxc5 7.Be4 f5 8.Kg2 Rg8 9.Rh1
9...Rg7 10.Bh6 Nd7 11.Bxg7+ Kxg7 12.Qxh7+ 1-0
In fact, Pierre is not sure this one is 100% correct. Any improvements
are wellcome!
# Position: 10
# Move: W
r.b..rk. Van der Wiel - Ribli Amsterdam 1980.
.p.nbppp Ribli is a solid player, but he didn't see this one
pq.p.... coming :
...B.... 15.Rxf7!! Rxf7 16.Bxf7 Kxf7 17.Qh5+ Kg8 16.Qe8+ Bf8
P..NP... 19.Nd5 Qxd4 20.Nxe7+ Kh8 21.Rf1 Qf6 22.Rxf6 and white on
..N.p... on move 29.
.PP...PP This is propably the easiest one, with number 5.
R..Q.R.K
# Position: 11
# Move: W
r.b...k. Teichmann - NN, Zurich 1920
p..p.nP. This was a trap, as the game was spectacularly won because
..pqr.Rp of black's weak play. Several authors have published this game
.p..p..P without checking anything. For instance, Neishtadt in his
..B.PnQ. excellent book "Lecons de tactique" (sorry, I don't know
.P...... the title of the english version) gives :
P.PP.... "1.Rxh6!! the g pawn will be helped by the h pawn 1...Nxh6
.K....R. (1...Rxh6 2.Bxf7+ Rxf7 3.g8Q+) 2.Qg5 Nf7 3.Qd8+!! Nxd8
4.h6 and there is nothing to prevent h6-h7+ 1-0"
Right, nothing can prevent h6-h7+, but it is not decisive!
4...Qd4!! 5.h7+ Kf7 6.g8Q+ Ke7 7.h8Q Kd6 8.Rg7 Qxd2!! 9.Qxd8 Rc5
10.Rxd7 Bxd7 11.Qxa8 Rb4 12.Qxa7 Qe1 13.Kb2 Qc3 1/2 as white cannot
escape the black checks.
-----------------------------------------------------------------------------
DT2 analysis:
----------------
>Sometime ago, the following 11 positions collected by Pierre Nolot were
>posted to rgc. Pierre claimed that no computers can solve any of them
>in reasonable time (and no micros in even a few weeks). A long while back, he
>might be right. At this point, a few of them are solvable even under
>tournament time control. The ones that we tried overnight are all solvable
>in a few hours. DT-2 was searching around 3 million nodes/sec during the
>runs. The time probably would be reduced by 5-30 times for the harder
>problems if the machine has a better quiescence search.
I just phoned Pierre, and he admits he is very impressed by DT-2'sresults!
So am I, and we thank you for the time you have spent on these positions.
I'll post the solutions according to Pierre as given in Gambisco
separately, but first I'll just give you his comments on DT's outstanding
results (much much better than the best micros)
>1 ;r3;q;b1;k/1;b4;p1/;p2;p;r2;p/3;n4/p;n;p1n1n1/6rp/1b3pp1/1b1qr1k1/40
>Source: Kasparov-Karpov, Lyon 1990.
>Kasparov played Nh6 here. DT-2 agreed with Kasparov's analysis after
>going down the line given in Informant, but could not find the move
>on its own in one hour time. When letting it run overnight, it produced
>Nh6 after 8 hours (but would play it with a 6 hours/move time control; that
>is, it got first indication that the combination might work after about 6
>hours), with the variation 1. Nh6! c3 2. Nf5 cb2 3. Qg4 ... which wins for
>white.
Very good. This is not an easy one!
>2 ;r4;r;k1/;p;p1;n1;p1;p/1;n;qP2;p1/2;b1p1b1/4nq2/1b3p2/pp2k2p/2r5/40
>Source: Bronstein-Ljubojevic, Petropolis 1973.
>Solution is 1. Rc5! Nc5 2. Nf6 Kh8 3. Qh4... Solved under tournament
>time control (3 minutes/move, plus extra time if the value for the move
>is unstable). Plays it after 2 minutes. Actually spent 4.5 minutes before
>the value stabilized to about a piece up.
Excellent!! and this is not supposed to be an easy one. Your score proves
you found it for the good reasons.
>3 ;r2;q;k2;r/;p;p;p1;b1;p;p/2;n1;p3/3;pp1;n1/3p2;b1/2pb1nn1/pp4pp/r1bqk2r/40
>Source: Smagin-Sahovic, Biel 1990.
>Solution move is 1. Ng5!. Verified that the move is sound by following
>the published analysis, but could not play it within one hour time. This
>is more a positional sac than what would be normally called tactics.
Well, Pierre still calls it tactics!
>4 ;r1;b1;k;b1;r/1;p1;n1;p;p;p/;p2;p;p;n2/6bb/2;qnp3/2n5/ppp2ppp/r2q1rk1/40
>Source: Keres-Kotov, Budapest 1950.
>Solution: 1. Ne6 Qe6 2. Nd5 Kd8 3. Bg4. Plays it after 3.8 hours, would play
>it under 2.5 hour/move time control.
Very good!
>5 ;r2;q;r;b1;k/1;p1;b2;p1/;p2;p;p;n1;p/8/3np3/1bn5/ppp3qp/1k3rr1/40
>Source: Spassky-Petrosian, Moskow 1969.
>Took 4 hours to find e5!. 1. e5! de5 2. Ne4 Nh5 3. Qg6... More than a
>piece up. The minimum time control needed to play it is around 2 hours/move.Very good, but 3.Qg4 is even stronger (spassky played Qg6)
Hiarcs also finds this one, and also gives 3Qg6
>6 ;r;n;b;q;k2;r/1;p3;p;p;p/;p7/1n;pp;p3/qpp1p1;n1/p4n2/4k;bpp/r1b2b1r/41
>Source: Melaniuk-Ivanchuk, USSR 1988.
>Ivanchuk played 1. ... ab5! which appears to be a sound positional sacrifice.
>Calling this position a tactical one, however, is a little unfair on Pierre's
>part. Ivanchuk, by the way, could not find the best continuations over the
>board, but won it anyway due to Melaniuk's inaccurate defence. Could not
>solve it in one hour. Probably a home preparation. The same move was
>replayed 3 years later in some random tournament.
All this is perfectly right, except that Pierre still thinks this is tactics.
Very deep tactics.
>7 1;r1;b;k2;r/2r2;p;p;p/;p3;p3/1;b2p2;q/4qp2/4n3/1b4pp/3r2k1/40
>8 - R - B K * - R
>7 * - r - * P P P
>6 P * - * P * - *
>5 * B * - p - * Q
>4 - * - * q p - *
>3 * - * - n - * -
>2 - b - * - * p p
>1 * - * r * - k -
> a b c d e f g h
>Source: unknown.
>This one we left the machine ran overnight. It found that 1. Rd8! wins.
>1. Rd8! Kd8 2. Ra7 and if 2. ... Qe2 3. Qd4 Ke8 4. h3... Took 12 hours
>to get the value back, but would have played it if the time control
>is 6 hours/move.
>8 ;r3;r;b;k1/;p;p;q2;p;p;p/2;b1;pb2/8/6q1/1p1b3p/p1p2pp1/r2r2k1/40
>White to move
>8 R * - * R B K *
>7 P P Q - * P P P
>6 - * B * P b - *
>5 * - * - * - * -
>4 - * - * - * q *
>3 * p * b * - * p
>2 p * p * - p p *
>1 r - * r * - k -
> a b c d e f g h
>Source: Gufeld-Osnos, USSR 1978.
>The move played was 1. Bh7, but as it turns out 1. Bh7 might not be the
>best move. Osnos defended poorly and got slaughtered. With best black
>defence, white only maintains a positional edge after 1. Bh7 Kh7 2. Qh5
>Kg8 3. Rd4 Bf3! (with the idea of Qxc2, and then Q to king side to defend
>the king). DT-2 prefers to play c4, which threatens Bh7 for real.
>9 ;r4;r1;k/4;b;p;p;b/2;n1;p2;p/;p1;n1p3/1;p1;p1bnp/3p1np1/;qp2qpb1/2rr2k1/40
>White to move
>8 R * - * - R - K
>7 * - * - B P P B
>6 - * N * P * - P
>5 P - N - p - * -
>4 - P - P - b n p
>3 * - * p * n p -
>2 Q p - * q p b *
>1 * - r r * - k -
> a b c d e f g h
>Source: unknown.
>After 9 minutes, it played 1. Nf6, expecting 1. ... Rfc8 2. Nh7 Kh7 3. Ra1 ...
>The score was slightly nagative for white. On longer searches, it went up
>to half a pawn and creeping up.
>10 ;r1;b2;r;k1/1;p1;n;b;p;p;p/;p;q1;p4/3b4/p2np3/2n1;p3/1pp3pp/r2q1r1k/40
>White to move
>8 R * B * - R K *
>7 * P * N B P P P
>6 P Q - P - * - *
>5 * - * b * - * -
>4 p * - n p * - *
>3 * - n - P - * -
>2 - p p * - * p p
>1 r - * q * r * k
> a b c d e f g h
>Source: Van der Wiel-Ribli, Amsterdam 1980.
>Solution is 1. Rf7 Rf7 2. Bf7 Kf7 3. Qh5. Plays it in less than 2 minutes,
>score stablized within 5 minutes to a pawn plus.
>11 ;r1;b3;k1/;p2;p1;np1/2;p;q;r1r;p/1;p2;p2p/2b1p;nq1/1p6/p1pp4/1k4r1/40
>White to move
>8 R * B * - * K *
>7 P - * P * N p -
>6 - * P Q R * r P
>5 * P * - P - * p
>4 - * b * p N q *
>3 * p * - * - * -
>2 p * p p - * - *
>1 * k * - * - r -
> a b c d e f g h
>Source: unknown
>Be6 appears somewhat worse for white, but 1. Rh6 Nh6 2. Qg5 Nf7 3. Qd8 appears
>to be a repetition draw or slightly ahead for white. Plays 1. Rh6 after
>5 minutes, thought white was slightly ahead, but going down the line, it
>appeared that black had enough pressure for a draw.
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.