mtp-50: a few more results

[jdart]

Arasan, 1 hr/move on a P4 3Ghz.

test 18 is not solved after an hour:
"18. Szabo-Sigurjonsson (Reykjavik) 1968" bm c5
result: Ba7 3600 seconds. score: +0.19 1064847003 nodes. ** error
Ba7 Nc5 Bxc5 bxc5 Qe7 Nb1 Ned7 Rc1 Ne4 Bb4 a5 Bxa5 Qxc5 Qb3 Qf5 Bb4
result(2): Bd6 3600 seconds. score: +0.19 1066137244 nodes. ** error
Bd6 Nc5 Bxc5 bxc5 Qe7 Nb1 Ned7 Rc1 Ne4 Bb4 a5 Bxa5 Qxc5 Qb3 Qf5 Bb4
result(3): Neg4 3600 seconds. score: +0.16 1056429328 nodes. ** error
Neg4 Nc5 Bxh2+ Kh1 Qc7 Bxg4 Nxg4 Qf5 Bg1 g3 Bxf2 Qxg4 Bxe1 Rxe1 Rcd8 Rd1 Re7 e4

test 20: Rxe6 is Arasan's third choice, but 3 moves are close in score:
"20. Olafsson-Levitt (Reykjavik) 1990" bm Rxe6
result: Qxc5 3600 seconds. score: +0.50 1240125271 nodes. ** error
Qxc5 Nc6 Qh5 h6 g4 Rab8 g5 Bg6 Qg4 h5 Qf4 Qb2 Rxd7 Qxe2 Rxd8+ Rxd8 Rxd8+ Nxd8 Qd6 Qd3
result(2): R1d3 3600 seconds. score: +0.31 1205000786 nodes. ** error
R1d3 Bc6 Qxc5 Qb1+ Bf1 Qxa2 Ne5 Qa4 Nxc6 Nxc6 Rxd7 Rxd7 Rxd7 Ne5 Rc7 Nd7 Qe7 Ne5
result(3): Rxe6 3600 seconds. score: +0.38 1291863034 nodes. ++ correct

test 23: g5 is Arasan's 3rd choice:
"23. Fischer-Larsen (Portoroz) 1958" bm g5
result: Bxf6 3600 seconds. score: +0.50 1169193067 nodes. ** error
Bxf6 Bxf6 Qf4 Qg5 Qxg5 Bxg5 Rh3 Rc5 Rf1 Bf4 Rfh1 e6 Ba4 Kf8 Ka1 Ke7 Rh7
result(2): Qd3 3600 seconds. score: +0.63 1210037989 nodes. ** error
Qd3 Qxd3 cxd3 Nd7 Be3 e6 Rh3 Nc5 Bc4 b3 axb3 Rb8 Bxc5 Rxc5 f4 a4 d4 Rc6
result(3): g5 3600 seconds. score: +0.31 1206485784 nodes. ++ correct
g5 Nh5 Bxg7 Kxg7 Qd4+ Kg8 Qd5 Qxd5 exd5 Rc5 Ba4 Kg7 Rh2 Nf4 Rdh1 Rg8 Rh7+ Kf8 R1h4 Nxd5 f4

test 30: 3 moves look similar in score:
"30. Spraggett-Vescovi (Moscow) 2004" bm Bxf5
result: Bh6 3600 seconds. score: +0.06 937847509 nodes. ** error
Bh6 Bf6 Bxf5 gxf5 Nxf5 Nc5 Nd2 Kh8 Qf3 h4 Nxd6
result(2): Bxf5 3600 seconds. score: +0.34 977108219 nodes. ++ correct
Bxf5 gxf5 Nxf5 Kh8 Nxe7 Rxe7 c5 dxc5 Bxc5 Ree8 Be7 Qb8 d6 Qa7 Nxe5
result(3): Nxh5 3600 seconds. score: +0.03 932809947 nodes. ** error

32. g4 is not the first choice:
"32. J.Polgar-Berkes (Budapest) 2003" bm g4
result: Bxa8 3600 seconds. score: -0.97 1109269663 nodes. ** error
Bxa8 g4 h4 Ba6 Be4 gxf3 Bxf3 Nf6 Kb1 Nd5 g3 Kg8 Rc1 b5 Qd3 Nb4 Qc3 Nd5
result(2): h4 3600 seconds. score: -0.97 1135300656 nodes. ** error
h4 g4 Bxa8 Ba6 Be4 gxf3 Bxf3 Nf6 Kb1 Nd5 g3 Kg8 Rc1 b5 Qd3 Nb4 Qc3 Nd5
result(3): g4 3600 seconds. score: -1.03 1108077402 nodes. ++ correct
g4 Rb8 h4 g6 hxg5+ Kg7 Qf4 Rh8 Rxh8 Qxh8 Qxc7 Qd8 Qf4 Bb7 Bxb7 Rxb7 d5 exd5 Qd4+ f6 Qxd5

36. exf6 looks pretty good, too:
"36. Morozevich-Ljubojevic (Monaco) 2002" bm g4
result: g4 3600 seconds. score: +1.50 1219364787 nodes. ++ correct
g4 Nc7 gxf5 exf5 Ngf3 a5 Rhg1 Ra8 Rg6
result(2): exf6 3600 seconds. score: +1.31 1259027254 nodes. ** error
exf6 Nxf6 Rde1 Rfe8 Qg6 Qf8 Ndxe6 Bxe6 Nxe6 Qf7
result(3): Ngf3 3600 seconds. score: +0.31 1264530304 nodes. ** error
Ngf3 Ra8 h5 a5 Nxb5 a4 Nfd4 axb3 Qxb3 Qc5 Rh3 Ra5 Rc3 Qb6 Na3 Qa7 Ndb5 Qa6

[Marc MP]

Thank you for the analysis. I'm currently working to improve the testsuite. I will try to make it more combinational (i.e. direct tactics). I'll check your suggestions.

Before you come to that note that: Qd1 wins as well on "47. Spassky-Portisch (Genève m/9) 1974"; and that "10. Rubesov-Borisenko (USSR) 1960"; just doesn't make sense (I found at least 2 refutations so that e5 is playable...).

Marc

[Marc MP]

Arasan, 1 hr/move on a P4 3Ghz.

test 18 is not solved after an hour:
"18. Szabo-Sigurjonsson (Reykjavik) 1968" bm c5
result: Ba7 3600 seconds. score: +0.19 1064847003 nodes. ** error
Ba7 Nc5 Bxc5 bxc5 Qe7 Nb1 Ned7 Rc1 Ne4 Bb4 a5 Bxa5 Qxc5 Qb3 Qf5 Bb4
result(2): Bd6 3600 seconds. score: +0.19 1066137244 nodes. ** error
Bd6 Nc5 Bxc5 bxc5 Qe7 Nb1 Ned7 Rc1 Ne4 Bb4 a5 Bxa5 Qxc5 Qb3 Qf5 Bb4
result(3): Neg4 3600 seconds. score: +0.16 1056429328 nodes. ** error
Neg4 Nc5 Bxh2+ Kh1 Qc7 Bxg4 Nxg4 Qf5 Bg1 g3 Bxf2 Qxg4 Bxe1 Rxe1 Rcd8 Rd1 Re7 e4

test 20: Rxe6 is Arasan's third choice, but 3 moves are close in score:
"20. Olafsson-Levitt (Reykjavik) 1990" bm Rxe6
result: Qxc5 3600 seconds. score: +0.50 1240125271 nodes. ** error
Qxc5 Nc6 Qh5 h6 g4 Rab8 g5 Bg6 Qg4 h5 Qf4 Qb2 Rxd7 Qxe2 Rxd8+ Rxd8 Rxd8+ Nxd8 Qd6 Qd3
result(2): R1d3 3600 seconds. score: +0.31 1205000786 nodes. ** error
R1d3 Bc6 Qxc5 Qb1+ Bf1 Qxa2 Ne5 Qa4 Nxc6 Nxc6 Rxd7 Rxd7 Rxd7 Ne5 Rc7 Nd7 Qe7 Ne5
result(3): Rxe6 3600 seconds. score: +0.38 1291863034 nodes. ++ correct

test 23: g5 is Arasan's 3rd choice:
"23. Fischer-Larsen (Portoroz) 1958" bm g5
result: Bxf6 3600 seconds. score: +0.50 1169193067 nodes. ** error
Bxf6 Bxf6 Qf4 Qg5 Qxg5 Bxg5 Rh3 Rc5 Rf1 Bf4 Rfh1 e6 Ba4 Kf8 Ka1 Ke7 Rh7
result(2): Qd3 3600 seconds. score: +0.63 1210037989 nodes. ** error
Qd3 Qxd3 cxd3 Nd7 Be3 e6 Rh3 Nc5 Bc4 b3 axb3 Rb8 Bxc5 Rxc5 f4 a4 d4 Rc6
result(3): g5 3600 seconds. score: +0.31 1206485784 nodes. ++ correct
g5 Nh5 Bxg7 Kxg7 Qd4+ Kg8 Qd5 Qxd5 exd5 Rc5 Ba4 Kg7 Rh2 Nf4 Rdh1 Rg8 Rh7+ Kf8 R1h4 Nxd5 f4

test 30: 3 moves look similar in score:
"30. Spraggett-Vescovi (Moscow) 2004" bm Bxf5
result: Bh6 3600 seconds. score: +0.06 937847509 nodes. ** error
Bh6 Bf6 Bxf5 gxf5 Nxf5 Nc5 Nd2 Kh8 Qf3 h4 Nxd6
result(2): Bxf5 3600 seconds. score: +0.34 977108219 nodes. ++ correct
Bxf5 gxf5 Nxf5 Kh8 Nxe7 Rxe7 c5 dxc5 Bxc5 Ree8 Be7 Qb8 d6 Qa7 Nxe5
result(3): Nxh5 3600 seconds. score: +0.03 932809947 nodes. ** error

32. g4 is not the first choice:
"32. J.Polgar-Berkes (Budapest) 2003" bm g4
result: Bxa8 3600 seconds. score: -0.97 1109269663 nodes. ** error
Bxa8 g4 h4 Ba6 Be4 gxf3 Bxf3 Nf6 Kb1 Nd5 g3 Kg8 Rc1 b5 Qd3 Nb4 Qc3 Nd5
result(2): h4 3600 seconds. score: -0.97 1135300656 nodes. ** error
h4 g4 Bxa8 Ba6 Be4 gxf3 Bxf3 Nf6 Kb1 Nd5 g3 Kg8 Rc1 b5 Qd3 Nb4 Qc3 Nd5
result(3): g4 3600 seconds. score: -1.03 1108077402 nodes. ++ correct
g4 Rb8 h4 g6 hxg5+ Kg7 Qf4 Rh8 Rxh8 Qxh8 Qxc7 Qd8 Qf4 Bb7 Bxb7 Rxb7 d5 exd5 Qd4+ f6 Qxd5

36. exf6 looks pretty good, too:
"36. Morozevich-Ljubojevic (Monaco) 2002" bm g4
result: g4 3600 seconds. score: +1.50 1219364787 nodes. ++ correct
g4 Nc7 gxf5 exf5 Ngf3 a5 Rhg1 Ra8 Rg6
result(2): exf6 3600 seconds. score: +1.31 1259027254 nodes. ** error
exf6 Nxf6 Rde1 Rfe8 Qg6 Qf8 Ndxe6 Bxe6 Nxe6 Qf7
result(3): Ngf3 3600 seconds. score: +0.31 1264530304 nodes. ** error
Ngf3 Ra8 h5 a5 Nxb5 a4 Nfd4 axb3 Qxb3 Qc5 Rh3 Ra5 Rc3 Qb6 Na3 Qa7 Ndb5 Qa6

Quickly:

"23. Fischer-Larsen (Portoroz) 1958" bm g5
g5 Nh5 Rxh5! wins

"32. J.Polgar-Berkes (Budapest) 2003" bm g4
This one is deep tactics I guess. After g4 Rb8 h4 Rybka 1.0 Beta (very tactical) thinks black is in trouble: 01:00:41.4 -1,27 17 133746283 g6 hxg5+ Kg7 Qf4 Bd6 Ne5. Considering that white has a bad game after Bxa8 g4, it seems that g4 is indeed best.

"36. Morozevich-Ljubojevic (Monaco) 2002" bm g4
Yes, ef6 is another good move.

[jdart]

My 1hr/move x 3 test run is almost complete now. Some more results:

37. White can play Qg8+ before playing exf6:
"37. Tal-Vooremaa (Tallinn) 1971" bm exf6
result: exf6 3600 seconds. score: +2.44 1189164514 nodes. ++ correct
exf6 Rxg7 fxg7 Kf7 f6 Bxf6 Nxh5 Ne7 Rae1 Ng8 Bh7 Nh6 Rxf6+ Qxf6 g8=Q+ Nxg8 Bxg8+ Kxg8 Nxf6+ Kf7 Ne4 Kg6 g4 b6
result(2): Qg8+ 3600 seconds. score: +1.94 1158332603 nodes. ** error
Qg8+ Rf8 Qg6+ Kd8 exf6 Bd6 Qg7 Qxb2 Rfe1 Qa3 Nd5 Ne5 f7 Rxf7 Qg8+ Rf8 Qg5+ Be7 Nxe7 Qxe7 Qxe7+ Kxe7 Rxe5+ Kd8 Rae1
result(3): Qh8+ 3600 seconds. score: +1.25 1133315838 nodes. ** error

42. I get 3 moves near equal in score:
"42. Nataf-Volokitin (?) 2004" bm Ng4
result: Bb7 3600 seconds. score: -0.13 1030684535 nodes. ** error
Bb7 Nd2 Nd7 Nc4 Be7 Qf3 g5 e5 Nxe5 Qg3 f6 Bxg5 O-O-O Bf4 Bd6 Nxd6+ Qxd6
result(2): a5 3600 seconds. score: -0.06 1036729291 nodes. ** error
a5 Nd2 Bb7 Nc4 Ba6 e5 Nd5 Rd1 Kf8 Qe4 Be7 Bf4 Qa7 Nd6 Rd8 Bxa6 Qxa6
result(3): Be7 3600 seconds. score: -0.13 1038771430 nodes. ** error
Be7 Nd2 Nd7 Nc4 Bb7 Qf3 g5 e5 Nxe5 Qg3 f6 Bxg5 O-O-O Bf4 Bd6 Nxd6+ Qxd6

Btw. event/site for this game is SCG-chT Budva.

44. Be4 is best but not by a lot:
"44. Yusupov-Enders (German Ch, Bremen) 1998" bm Be4
result: Be4 3600 seconds. score: +0.63 1135706489 nodes. ++ correct
Be4 Qb5 Bxb1 Qxb1 Rd7 Qb2 Rdc7 Bd2 Qd5 Qb1 Qa5 Rg3 Kh8 Qb4 Qa6 Nf3 N8d7
result(2): Nd5 3600 seconds. score: +0.33 1075086623 nodes. ** error
Nd5 Qd2 Qc7 Bb2 Nd7 Rg3 N7f6 c4 Nf4 Bc1 N4h5 Rc3 Qd6
result(3): Ne4 3600 seconds. score: +0.19 1081917563 nodes. ** error

47. you said was broken - I agree.

"48. Geller-Dreev (New York) 1990" bm Bxc4
I get Bxc4 but not with a very convincing score:
result: Bxc4 3600 seconds. score: +0.34 1072612820 nodes. ++ correct
Bxc4 dxc4 Nxc4 Be7 Rb1 Ra7 Bf4 Kf8 Re1 Nd7 d5 Qa8 d6 Bf6 Qe2 Bc3 Bd2 Bxf3
result(2): Rb1 3600 seconds. score: -0.13 1025710032 nodes. ** error
Rb1 Bc6 Ne5 Bxe5 dxe5 Bxa4 Bg4 c3 Nf3 h5 Bh3 Bb5 Re1 Ne7 Bg5 Nbc6
result(3): Nxc4 3600 seconds. score: -0.09 1046202605 nodes. ** error
Nxc4 dxc4 Bxc4 Qc7 Qe2+ Kf8 Ng5 Qc6 f3 Nh6 Re1 Nd7 Nxf7 Nxf7 Qe6 Bxh2+ Kxh2 Qxe6 Bxe6

Overall a number of the positions are really good, but I tend not to trust testsuites until I have done this kind of analysis. I think test positions should be hard enough to be interesting but not so hard to require an hour+ to solve. (Although in a few years, with hardware and software improvements, these may be not too hard!).

[jdart]

49. is also not solved after an hour:
"49. Smirnov-Kurnosov (Istanbul) 2004" bm Bxh5
result: Qf2 3600 seconds. score: +0.19 1034198113 nodes. ** error
Qf2 Ne7 Nd2 Bxc3 bxc3 d5 Bd4 Bd7 Rd1 Qb7 Re3 Rec8 exd5 Qxd5 Ne4
result(2): Bf2 3600 seconds. score: +0.13 1041022545 nodes. ** error
Bf2 d5 e5 Nd7 Qg3 Bb7 Nb1 Qc8 Qf3 Qc7 N1d2 Kh8 c3 Nb6 Bc5
result(3): Rd1 3600 seconds. score: +0.06 1034777037 nodes. ** error