Author: Christophe Theron
Date: 22:57:50 11/18/99
Go up one level in this thread
On November 18, 1999 at 23:34:02, Peter McKenzie wrote:
>Thanks for an interesting post. I always like it when people put some thought
>into analysing actual chess positions and the play that results from them.
>
>Christophe, how about posting scores from Tiger?
OK, I post the output of the REBEL-TIGER inside the text. It has the 12.0
engine.
I have used a K6-2 300MHz with 8Mb hash tables. I have nothing faster.
Divide the time by 1.5 to get the times on the hardware used by the SSDF.
I don't give comments, as I did not have enough time to have a deep look at the
positions.
>I have taken a bried look at these positions and quickly run them on LambChop
>for about 60sec on a PII 366mhz. Comments below:
>
>On November 18, 1999 at 17:33:57, Howard Exner wrote:
>
>>Here are a few examples of nice moves made by Tiger. Perhaps other programs make
>>these moves also.
>>
>>1. Of course in this example white can play anything and win but Tiger chooses
>>Qxf7.
>>Here Tiger readily enters an endgame with the move Qxf7. This is exactly how any
>>strong human would play. Most programs would most likely move the Queen. It is
>>for me at least, an indicator that something unique is going on with Tiger's
>>endgame eval. If it plays a move like Qxf7
>>maybe it knows something about liquidating into completely won endgames.
>>
>>8/p2k1r2/1p2bQ2/3p4/8/1P4P1/1PP2R1P/6K1 w - - id Chess Tiger - Goldbar; bm f6f7;
>
>Yes, Qxf7 is the human move and a nice way to tidy up the game. My program kept
>switching between Qxf7, Qd4 and Qe5, in the end choosing Qe5 with a +10.9 score.
>Would be nice if it played Qxf7, but I'm not too worried when it sees such a big
>score. Food for though though.
00:00:00.636 11,06 7 Qd4 Kd6 Qb4+ Kc6 Rxf7 Bxf7 Qf8 Be6
00:00:01.232 11,48 7 Qd4 Rxf2 Kxf2 Kd6 Qb4+ Kc7 Qe7+ Bd7 h4 a6
00:00:02.115 11,52 8 Qd4 Rxf2 Kxf2 Kc6 Qa4+ Kb7 g4 a6 h4 Kc7
00:00:06.750 11,78 9 Qd4 Rxf2 Kxf2 Kc6 Qa4+ Kb7 g4 a5 h4 Kc7 h5
00:00:15.105 11,84 10 Qd4 Rxf2 Kxf2 Kc6 Qa4+ Kb7 g4 a6 h4 b5 Qd4 Kc6
00:00:19.759 11,85 10 Qe5
00:00:38.064 12,23 10 Qe5 Rxf2 Kxf2 Bf7 Qb8 Be6 Qxa7+ Kc6 Qa8+ Kd6 Qb7
Bf5
00:00:41.740 12,24 10 Qxf7+
00:00:41.910 12,68 10 Qxf7+
00:00:44.344 15,08 10 Qxf7+ Bxf7 Rxf7+ Ke6 Rf1 a5 h4 a4 h5 axb3 h6
bxc2 h7 d4 g4
00:00:48.478 15,08 11 Qxf7+ Bxf7 Rxf7+ Ke6 Rf1 a5
00:00:49.442 15,98 12 Qxf7+
>>2. This game illustrates an excellent liquidation into a won endgame. Noteworthy
>>are the
>>two rook captures on f6. The move Bh3 is one of the nicest examples I've seen of
>>a program
>>swapping down into an easily won endgame. Here are the three moves and the
>>entire game.
>>
>>1rb2rk1/p1q2ppp/2nb1n2/2p1p3/4P3/1PNPN1P1/1BPQ2BP/R4RK1 w - - id Chess Tiger
>>12.0 K6 - Nimzo '99 P200MMX; bm f1f6;
>
>Initially Chop liked Nd5, but changed to Rxf6 after 6sec. Stayed with that move
>for the 60sec. Score +.98 pawns after 9ply search.
00:00:00.507 1,28 6 Nf5 Be7 Qg5 Bxf5 exf5 Nd4
00:00:00.892 1,28 7 Nf5 Be7 Qg5 Bxf5 exf5 Nd4 Rf2
00:00:02.138 1,18 7 Nf5 Be7 Nxe7+ Nxe7 h3 Qb6 Qf2 Bb7 Bf3
00:00:05.690 1,20 8 Nf5 Be7 Nxe7+ Qxe7 Nd5 Nxd5 exd5 Nd4 d6 Qd7 Ra5
00:00:19.362 1,31 9 Nf5 Bxf5 exf5 Qb6 Qf2 Nb4 Kh1 Qc7
00:00:41.613 1,28 10 Nf5 Bxf5 exf5 Nd4 Nd5 Nxd5 Bxd5 Be7 Qf2 Rfd8 Bc4
Rd6
00:01:40.217 1,27 11 Nf5 Bxf5 exf5 c4
00:04:55.939 1,26 12 Nf5 Bxf5 exf5 Nd4 Nd5 Nxd5 Bxd5 Be7 Ra6 Rfd8 Be4
f6
00:05:50.420 1,27 12 Rxf6
00:06:00.580 1,50 12 Rxf6 gxf6 Ned5 Qd8 Qh6 Be7 Nxe7+ Nxe7 Qxf6 Qd4+
Kh1 Ng6 Rxa7 c4 Ra4
>
>>1r1q1rk1/p3bp1p/2n1bp1Q/2pNp3/4P3/1PNP2P1/1BP3BP/5RK1 w - - id Chess Tiger 12.0
>>K6 - Nimzo '99 P200MMX; bm f1f6;
>
>Switching between Nxf6 and Rxf6 for a while. Locked onto Rxf6 after 7sec.
>Score +1.31 after 8ply.
00:00:00.473 1,02 5 Nxe7+ Nxe7 Rxf6 Qd4+ Kh1 Kh8 Bf3
00:00:00.719 1,14 5 Nxe7+ Nxe7 Na4 Qd6 Rxf6 Kh8 Bf3
00:00:01.656 1,08 6 Nxe7+ Nxe7 Qxf6 Nc6 Nd5 Qxf6 Nxf6+ Kh8 Rf3
00:00:04.242 1,11 7 Nxe7+ Nxe7 Na4 Ng6 Qe3 Kh8
00:00:10.549 1,08 8 Nxe7+ Nxe7 Na4 c4 dxc4 Ng6 Qe3 Kg7
00:00:24.249 1,08 9 Nxe7+ Nxe7 Na4 c4 dxc4 Ng6 Qe3 a5
00:01:28.725 1,12 10 Nxe7+ Nxe7 Nd5 Bxd5 exd5 Ng6 Rf5 Qa5 Rxf6 Qe1+
Rf1 Qe2 Rc1
00:01:50.364 1,13 10 Rxf6
00:03:07.730 1,79 10 Rxf6 Rc8 Nxe7+
>>2r2rk1/p2qbp1p/2n1bR1Q/1NpNp3/4P3/1P1P2P1/1BP3BP/6K1 w - - id Chess Tiger 12.0
>>K6 - Nimzo '99 P200MMX; bm g2h3;
>
>I don't get Bh3 here, I'll have to take a longer look at it when I get home.
>I stay with Qg5+ from ply 2.
>I let it run for couple of min, the score was +3.74 after 9ply.
>PV starts with 1.Qg5+ Kh8 2.Bxe5+ Nxe5 3.Qxe5+ Kg8 4.Bh3
>Does this transpose?
>Perhaps an improvement over Bh3 !?
00:00:00.593 2,10 6 Qg5+ Kh8 Bxe5 Nxe5 Nxe7 Qxe7 Qxe5 Kg8 Nd6
00:00:01.494 2,23 7 Qg5+ Kh8 Bxe5 Nxe5 Qxe5
00:00:03.064 2,38 8 Qg5+ Kh8 Bxe5 Nxe5 Qxe5 Kg8 Bh3 Rcd8 Bxe6 fxe6
Nxe7+ Qxe7 Rxe6
00:00:07.629 2,36 9 Qg5+ Kh8 Bxe5 Nxe5 Qxe5 Kg8 Bh3 Rc6 Bf5 Ra6 Rh6
f6 Nxe7+ Qxe7
00:00:14.909 2,37 9 Bh3
00:00:19.603 2,88 9 Bh3 Bxh3 Qg5+ Kh8 Bxe5 Nxe5 Qxe5 Qxd5 exd5 Bxf6
Qxf6+ Kg8 Nxa7
00:00:28.775 2,88 10 Bh3 Bxh3 Qg5+ Kh8 Bxe5 Nxe5 Qxe5 Qxd5 exd5 Bxf6
Qxf6+ Kg8 Nxa7
00:01:16.080 3,12 11 Bh3 Bxh3 Qg5+ Kh8 Bxe5 Nxe5 Qxe5 Bxf6 Qxf6+ Kg8
Ne7+ Qxe7 Qxe7 a6 Nd6
Tiger has the same PV as LambChop as after 3 seconds, but apparently playing Bh3
first does not transpose into the same thing.
>>[Event "SSDF Game 40/2"]
>>[Site "Hans C."]
>>[Date "1999.11.07"]
>>[Round "22"]
>>[White "Chess Tiger 12.0 K6450"]
>>[Black "Nimzo '99 P200MMX"]
>>[Result "1-0"]
>>[WhiteElo "?"]
>>[BlackElo "?"]
>>[ECO "B24"]
>>
>>1. e4 c5 2. Nc3 Nc6 3. g3 Rb8 4. Bg2 b5 5. Nge2 b4 6. Nd5 e6 7. Ne3 Nf6
>>8. d3 Be7 9. a3 bxa3 10. b3 O-O 11. O-O d6 12. Bxa3 e5 13. Bb2 Qc7 14.
>>f4 Bd7 15. fxe5 dxe5 16. Nc3 Bd6 17. Qd2 Bc8 18. Rxf6 gxf6 19. Ned5 Qd8
>>20. Qh6 Be7 21. Rf1 Be6 22. Rxf6 Rc8 23. Nb5 Qd7 24. Bh3 Bxh3 25. Qg5+
>>Kh8 26. Bxe5 Nxe5 27. Qxe5 Bxf6 28. Qxf6+ Kg8 29. Ne7+ Qxe7 30. Qxe7 f6
>>31. Nxa7 Rf7 32. Nxc8 Rxe7 33. Nxe7+ Kf7 34. Nd5 Be6 35. Kf2 f5 36. Ke3
>>h6 37. Nf4 Bd7 38. e5 Bc6 39. e6+ Ke7 40. c4 1-0
>>
>>3. A nice move exploiting some of black's kingside weaknesses.
>>
>>1rb1r1k1/pp3p1p/2np2p1/q1p5/3bPB2/2NP1BP1/PPPQ1R1P/5RK1 w - - id Chess Tiger
>>12.0 K6 - Nimzo '99 P200MMX; bm f4d6;
>
>Chop locks on to Bxd6 straight away, and holds for 60sec.
>Score is only about +.26 though.
00:00:00.422 0,56 6 Bxd6 Ra8 Ra1
00:00:00.590 0,40 6 Bxd6 Ra8 Kg2 Bxf2 Qxf2 Nd4 Bd1
00:00:01.300 0,36 7 Bxd6 Ra8 Kh1 Bh3 Bg2 Bxf2 Qxf2 Bxg2+ Kxg2
00:00:03.389 0,34 8 Bxd6 Ra8 Kg2 Bxf2 Qxf2 c4 Nd5 Re6 Ba3 cxd3 cxd3
00:00:07.069 0,36 9 Bxd6 Ra8 Kg2 Bxf2 Qxf2 c4 Nd5 Re6 Ba3 cxd3 cxd3
00:00:23.259 0,38 10 Bxd6 Ra8 Kg2 Be6 Re2 c4 e5 cxd3 cxd3 Rac8 Nd5
Qxd2
00:01:03.845 0,15 11 Bxd6 Ra8 Kg2 Qd8 Bf4
00:02:15.737 0,15 12 Bxd6 Ra8 Kg2
>>4. Nxc5 leads to active piece play for white. The next few moves went 1 ... Nxf1
>>2. Kxf1 Bc8
>>and black's position is extremely cramped. Whether or not Nxc5 is the best move
>>is debatable(black's move Bc8 may not be the best defense)but it appears to
>>reveal an awareness of how the relative value of pieces change depending on
>>position.
>>
>>r2q1rk1/pp1b1pbp/2n1p1p1/2p1P3/P2pNP2/1P1PnNP1/2PQ2BP/R1B2RK1 w - - id Chess
>>Tiger 12.0 do - Shredder 2 P200MMX ; bm e4c5;
>
>Interesting. Chop wants to move its rook, switching between Re1 and Rf2 with a
>slight plus to white. I have a later version at home with better outpost
>scoring that might choose differently here. I doubt it will sac the change
>though.
00:00:00.760 0,22 6 Rf2 b6 Qe2 Nxg2 Rxg2 Nb4
00:00:00.982 0,20 6 Rf2 Nxg2 Rxg2 b6 Qe1 Ne7
00:00:02.018 0,18 7 Rf2 b6
00:00:04.251 0,13 8 Rf2 b6 Nd6 Nxg2 Rxg2 f6 exf6 Bxf6 Ng5
00:00:19.586 0,00 9 Rf2 Nxg2 Kxg2 b6
00:00:38.263 0,01 10 Rf2 Nxg2 Rxg2
00:01:16.429 -0,08 11 Rf2 Nxg2 Kxg2 b6 Qe1 a5 Nd6 Nb4 Nb7 Qc7 Nd6 Bc6
00:02:11.932 -0,07 11 Nxc5
00:04:03.405 0,02 12 Nxc5 b6 Ne4 Nxf1 Kxf1 f6 exf6 Bxf6 Nxf6+ Qxf6 Ba3
Rfd8
Christophe
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.