new test positions

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
Eelco de Groot
Posts: 4564
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Re: new test positions

Post by Eelco de Groot »

In the last two iterations Blueberry has gone back to 28. Ncd4


17/63 229:23 -0.79 27...Re8 28.Ncd4 Bh3 29.Rd1 Nb4
30.Bf3 Bd7 31.Nc7 Bxd6 32.Nxe8 Kxe8
33.Nf5 Bxf5 34.Rxd6 Ke7 35.Rd1 a6
36.Ra1 Bd7 37.Re1+ Kd6 38.Be4 Be6
39.Rd1+ Kc5 40.Rc1+ Kd4 (9.190.579.619) 667

18/63 332:23 -0.82 27...Re8 28.Ncd4 Bh3 29.Rd1 Nb4
30.Bf3 Bd7 31.f5 a6 32.Nc7 Bxd6
33.Nde6+ Ke7 34.Nxe8 Bxe8 35.Nxg7 Bb5
36.Ne6 Be5 37.Nd4 Bd7 38.Kg2 Kf8
39.Ne6+ Ke7 40.Rc1 (13.653.784.266) 684


Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
User avatar
Eelco de Groot
Posts: 4564
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Re: new test positions

Post by Eelco de Groot »

I've got two more iterations, No. 20 is just in. It may be that 27... Re8 actually draws! The evaluation seems to stabilize. Jon, no need of course to actually include this test-position, as you can see from the number 001 I plan to make some sort of testsuite with similar positions. But up to you if you want to use it.

I lose a lot of testgames against Rybka 2.2n2 so there is no scarcity of testpositions 8-) I think I would need at least a 100 of these...


19/63 635:24 -0.69 27...Re8 28.Ncd4 Bh3 29.Rd1 Nb4
30.Bf3 Bd7 31.Nc7 Bxd6 32.Nxe8 Kxe8
33.Nf5 Bxf5 34.Rxd6 Ke7 35.Rd1 a6
36.Re1+ Kd6 37.Be4 Be6 38.Rd1+ Kc5
39.Rc1+ Kd4 40.Bxh7 Bxb3 (26.568.446.056) 696

20/63 950:42 -0.69 27...Re8 28.Ncd4 Bh3 29.Rd1 Nb4
30.Bf3 Bd7 31.f5 a6 32.Ne6+ Bxe6
33.fxe6 axb5 34.e7+ Rxe7 35.dxe7+ Kxe7
36.Kg2 Bd6 37.Be2 f5 38.Bxb5 g5
39.Re1+ Kf6 40.h4 g4 (39.443.250.199) 691

Regards, Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
jdart
Posts: 4366
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: new test positions

Post by jdart »

I ran # 1 with Rybka3 on a fast quad and got the following results - strangely it never wanted to play Rb2+:

Fen: 2r3k1/2qn1pPp/3n4/3Pp1P1/PQ4R1/1P6/2r1N2P/1K1R1B2 b - a3 0 1
Bm: Rxe2

yes 2 0 -8 443 Rxe2
yes 3 0 -8 637 Rxe2
yes 4 20 +6 1172 Rxe2
yes 5 20 +1 2278 Rxe2 Bxe2 Qc2+
yes 6 30 +21 4228 Rxe2
yes 6 50 +34 6379 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf3
yes 7 140 +39 17465 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Rgg1 Nc5 Rc1
yes 8 200 +43 26086 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Rgg1 Nc5 Rb1 Kxg7
yes 9 520 +37 66441 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Rgg1 Nc5 Rc1 Qe3
Rb1 Kxg7 h4
no 9 800 +38 101512 Rb8 Qa3 Nc4
no 10 1610 +35 212313 Rb8 Qa3 Nc4 Rxc4 Rxc4 Rc1 Qc5 Qxc5 Rxc5
Ka2 Rxd5 Ka3 Kxg7 Ng3 Kg6
yes 10 1800 +37 239836 Rxe2 Bxe2 Qc2+
yes 11 2020 +37 270664 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Rgg1 Nc5 Rc1 Qe3
Rb1 Kxg7 h4
yes 12 3190 +57 431216 Rxe2
yes 12 3780 +77 523558 Rxe2
yes 12 7140 +101 1014874 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf3 Rg3 Qxd5
Rh3 e4 Rf1 Kxg7 Qa3 Ne5
yes 13 9780 +121 1440489 Rxe2
yes 13 14500 +141 2217815 Rxe2
yes 13 22450 +165 3537618 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf3 Rg3 Qxd5
Rh3 e4 g6 fxg6 Rc3 Qe5 Rc1 Kxg7 Ka2 Rb8
yes 14 35340 +185 5710436 Rxe2
yes 14 48640 +164 7937036 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Rgg1 Rc2 Qa3 Qf3
Qb4 Nc5
yes 15 59720 +184 9781682 Rxe2
yes 15 90160 +204 15095540 Rxe2
yes 15 157360 +206 26579830 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf3 Rg3 Qxd5
Rf1 Kxg7 Rh3 e4 Ka2 e3 Re1
yes 16 212140 +219 35810779 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Rdg1 Nc5 R4g2 Qf3
Ka2 Qxd5 Rc2 Kxg7 Rd1 Nd3 Qd2 Ne4 Qxd3
Qxd3 Rxd3 Rxc2+ Kb1 Rxh2 a5 h6 gxh6+ Rxh6
b4 Ra6 Rd7 Kf6
yes 17 322750 +219 53258248 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf2 Reg1 Nc5
R4g2 Qf3
yes 18 350950 +219 58110754 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf2 Reg1 Nc5
R4g2 Qf3
yes 19 498690 +239 82842867 Rxe2
yes 19 664240 +250 112888414 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf2 Reg1 Nc5
R4g2 Qf3
yes 20 1139100 +270 194957319 Rxe2
yes 20 1595220 +290 277824430 Rxe2
yes 20 2250630 +302 394825268 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf2 Reg1 Nc5
R4g2 Qf3
yes 21 3630110 +321 641375389 Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf2 Reg1 Nc5
R4g2 Qf3

Result: Success
Found in: 1800 ms (00:00:01.800)

-------------------------------------------------------------------------------

Results:
Top
User avatar
Eelco de Groot
Posts: 4564
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Re: new test positions

Post by Eelco de Groot »

+321 centipawns for Black when Black is several pawns down seems a bit optimistic from Rybka do I read that correctly? Otherwise I don't understand why Rybka switches to Rb8 with higher score. Blueberry thinks it is more even. The calculation of the PVs is going very fast so I do not entirely trust that it is correct, it's been known to fail around 0.00. But I trust the results for the other alternative root-moves as being worse than α if α = 0, better.

2r3k1/2qn1pPp/3n4/3Pp1P1/PQ4R1/1P6/2r1N2P/1K1R1B2 b - a3 bm Rxe2; id "arasan11.201";

[D]2r3k1/2qn1pPp/3n4/3Pp1P1/PQ4R1/1P6/2r1N2P/1K1R1B2 b - -

Engine: Blueberry Beta 4 DM70 Build 379 (Athlon 2009 MHz, 256 MB)
by F. Letouzey, T. Gaksch, E. de Groot

1/11 0:00 -2.26 1...Kxg7 2.Ka1 Kh8 (326)

2/17 0:00 -2.26 1...Kxg7 2.Ka1 Kh8 (4.330)

2/30 0:00 -2.20 1...Ra8 2.Rd2 Rxd2 3.Qxd2 Qb6 4.Qc3 Kxg7
5.h4 Rc8 6.Qd3 (180.800)

2/33 0:00 -1.96 1...Ra8 2.Rd2 Rc8 3.Rg2 Kxg7 4.h4 Qc5
5.Qxc5 R2xc5 6.Rg3 Nb6 (403.623)

2/42 0:02 -1.96 1...Ra8 2.Rd2 Rc8 (1.169.527) 486

2/42 0:02 -1.43 1...Nb6 2.Rg2 Rb8 3.Rg3 Rb2+ 4.Ka1 Nbc4
5.g6 hxg6 (1.340.248) 486

2/47 0:08 0.00 1...Nb6 2.Qe1 Qc5 3.Qb4 Qc7 (4.229.245) 481

2/47 0:11 0.00 1...Nb6 2.Qe1 Rxe2 3.Qxe2 Nxd5
4.Rxd5 Qc1+ 5.Ka2 Rc2+ 6.Qxc2 Qxc2+
7.Ka3 Kxg7 8.Rxd6 Qc5+ 9.b4 Qxd6
10.a5 Qd2 11.Rc4 Kh8 12.a6 Qe3+
13.Ka2 Qf2+ 14.Ka3 Qe3+ (5.950.997) 497

3/47 0:25 -2.02 1...Nb6 2.Ka1 Nxd5 3.Qa3 (12.972.168) 511

3/47 0:27 -1.96 1...Ra8 2.Rd2 Rc8 3.Rg2 Kxg7 4.h4 Qc5
5.Qxc5 R2xc5 6.Rg3 Nb6 (13.779.585) 507

3/47 0:28 -1.96 1...Ra8 2.Rg3 Kxg7 3.Rc1 Qb7 4.Qxb7 Rxc1+
5.Nxc1 Nxb7 6.Kb2 e4 7.Ka3 (14.445.275) 502

3/50 0:34 -1.96 1...Ra8 2.Rg3 Rb2+ 3.Kxb2 Nc5 4.Rc3 Nde4
5.Ka3 Nxc3 6.Nxc3 (17.098.302) 498

3/50 0:37 -1.09 1...Kxg7 2.Ka1 Kh8 3.Rg3 f5 4.Qa3 Ne4
5.Rf3 Qb7 6.Rxf5 Nd2 7.Rxd2 Rxd2 (18.492.802) 496

3/50 0:38 -1.09 1...Kxg7 2.Ka1 Kh8 3.Qa3 f5 4.Rg1 Nb5
5.axb5 (19.146.156) 497

3/54 1:20 -0.08 1...Kxg7 2.Ka1 Nf5 3.Rc4 Rxc4 4.bxc4 Qxc4
5.Qxc4 Rxc4 6.Bh3 Rxa4+ 7.Kb1 Kg6
8.Ng3 Rb4+ 9.Ka2 Ra4+ 10.Kb2 Rb4+
11.Kc3 (40.966.250) 506

3/54 1:24 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Ra2+ (42.752.584) 506

3/54 1:39 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Qxg2 8.Rxg2 Rxg2 9.Qxd6 Rg1+
10.Kb2 Rg2+ 11.Ka1 (51.202.175) 515

3/54 2:01 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Qxg2 8.Rxg2 Rxg2 9.Qxd6 Rg1+
10.Kb2 Rg2+ 11.Ka1 (63.821.213) 525

4/54 2:22 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Ra2+ (74.968.582) 524

5/54 3:17 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Ra2+ (104.131.956) 528

6/54 5:38 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Ra2+ (181.187.927) 534

7/57 24:09 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Ra2+ (762.275.002) 525

8/58 41:05 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Ra2+ (1.300.117.180) 527


Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
jdart
Posts: 4366
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: new test positions

Post by jdart »

I think if you are getting a zero score something is wrong.

After .. Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Rdg1 Rybka selects .. Nc5:

[D] 2r3k1/5pPp/3n4/2nPp1P1/PQ4R1/1P6/4q2P/K5R1 w - - 0 5

It looks like Black can pick up a couple of pawns and in addition still has threats against the White king. One line from here is .. Nc5 R4g2 Qd3 (Qf3 is good too) Rb1 Qxd5 Rd2 Nd3 Qb6 Ne4 and Black is winning.
Top
jdart
Posts: 4366
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: new test positions

Post by jdart »

Well, maybe this is not so clear and White can wriggle out of this. This is Arasan's output at 1hr/move:

bm Rxe2
result: Rxe2 score: +0.68 ++ solved in 1363.50 sec. (3261.89M node
s)
Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf3 Rg3 Qxd5 Rh3 e4 Qa3 Qd2 Rhe3 Qxh2 Rd1 Kxg7 Qb2+
Qxb2+ Kxb2 Rc6 Rc3 Nc5 Rxd6 Nxa4+
result(2): e4 score: +0.55 ** not solved in 3600.02 secs. (
8628.19M nodes)
e4 Qa3 Ne5 Rh4
result(3): Rb2+ score: +0.49 ** not solved in 3600.01 secs. (
8651.01M nodes)
Rb2+ Kxb2 Qc2+ Ka3 Qxd1 Qxd6 Rb8 b4 Rc8 Qc6 Rxc6 dxc6 Nb6 b5 e4 Rxe4
Top
User avatar
Eelco de Groot
Posts: 4564
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Blueberry Beta 4 Build 392 result arasan 11.201

Post by Eelco de Groot »

jdart wrote:Well, maybe this is not so clear and White can wriggle out of this. This is Arasan's output at 1hr/move:

bm Rxe2
result: Rxe2 score: +0.68 ++ solved in 1363.50 sec. (3261.89M node
s)
Rxe2 Bxe2 Qc2+ Ka1 Qxe2 Re1 Qf3 Rg3 Qxd5 Rh3 e4 Qa3 Qd2 Rhe3 Qxh2 Rd1 Kxg7 Qb2+
Qxb2+ Kxb2 Rc6 Rc3 Nc5 Rxd6 Nxa4+
result(2): e4 score: +0.55 ** not solved in 3600.02 secs. (
8628.19M nodes)
e4 Qa3 Ne5 Rh4
result(3): Rb2+ score: +0.49 ** not solved in 3600.01 secs. (
8651.01M nodes)
Rb2+ Kxb2 Qc2+ Ka3 Qxd1 Qxd6 Rb8 b4 Rc8 Qc6 Rxc6 dxc6 Nb6 b5 e4 Rxe4

Fascinating stuff Jon!

And thanks very much for your help Jon!

Regarding Blueberry, I have some ideas to improve the output compared to build 379 but those are not yet fully implemented.

Especially the shorter PV searches can be a weakness for Blueberry and cause it to lose the thread. Whenever there has been a change at the root, the new bestmove has already been searched to greater depth than the iteration counter suggests.

All non-PV rootmoves are at present researched to nominal depth + 7 with a null window search, followed by a few PV searches if the move is > α. (In the terminology of Gian-Carlo you could look at this as hammering on all the other moves, but now hammering upwards until one is higher than the PV. And this of course is in a non-cluster situation. There is some relation though I think).
But the PV move itself at the start of the new iteration is in a worst case then only (re-)searched to the nominal depth and sometimes the evaluation drops far too much and the quality of PV and search suffers. This I have to improve upon somehow.
Well, maybe this is not so clear and White can wriggle out of this.
There would be some satisfaction in succeeding to prove Rybka wrong :P But Rybka plays the move right from the start and does not waver, that is generally a sign that she understands the position well. Blueberry does get a slightly rising score at the moment, following my last post the analysis with build 379 continued, Rxe2 was lost for a while sadly, regained then the evaluation went up again.

8/58 41:05 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Ka1 Ra2+ (1.300.117.180) 527

8/58 84:55 +0.63 1...e4 2.Rg3 Ne5 3.Qa3 Nf3 4.Ka1 Kxg7
5.Kb1 Nd2+ 6.Rxd2 Rxd2 7.Rc3 Qd8
8.Qc1 Rxd5 9.h4 Rd3 10.b4 Qd7
11.Rxc8 Qxc8 12.Qb2+ Kg8 (2.654.281.856) 520

8/62 110:46 +0.63 1...e4 2.Rg3 Ne5 3.Nc3 Rxc3 4.Rxc3 Qxc3
5.Qxc3 Rxc3 6.Kb2 Rf3 7.h4 Kxg7 8.a5 Rf2+
9.Kc3 f5 10.Kd4 Ng6 11.a6 f4 12.Ra1 Rd2+
13.Kc3 Rxd5 (3.360.585.599) 505

8/62 119:34 +0.63 1...e4 2.Rg3 Ne5 3.Nc3 Rb2+ 4.Kxb2 Nec4+
5.bxc4 Nf5 6.Nxe4 Qe5+ 7.Kc2 Rb8
8.d6 Nd4+ 9.Kc1 Rxb4 10.d7 Nb3+
11.Kc2 Na1+ 12.Kd2 Qb2+ 13.Kd3 Rb3+
14.Nc3 Qc2+ (3.647.452.192) 508

9/62 120:24 +0.08 1...e4 2.Rg3 Ne5 3.Qa3 Nf3 4.Ka1 Kxg7
5.Kb1 Nd2+ 6.Rxd2 Rxd2 7.Rc3 Qb6
8.Rxc8 Nxc8 (3.676.077.718) 508

9/62 132:04 +1.16 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rge1 Kxg7 7.a5 Nf5
8.Red1 Ne3 9.Rb2 Qg3 10.Qc3 Qxg5
11.Re1 Nxd5 12.Rxe5 Nxc3 13.Rxg5+ Kf6
14.Rg4 (4.006.858.118) 505

9/62 135:43 +1.16 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rge1 Kxg7 7.a5 Nf5
8.Red1 Ne3 9.Rb2 Qc2 10.Rd2 Nxb3+
11.Rxb3 Qc1+ 12.Ka2 Nc2 13.Rxc2 Rxc2+
14.Rb2 Rc4 (4.124.354.643) 506

9/62 196:21 +2.62 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Re1 Qf3 5.Rg3 Qxd5 6.h4 Rb8 7.Qg4 Nc5
8.Qd1 Qe6 9.Ree3 Nf5 10.Qg4 Rb6
11.Ref3 Nxg3 12.Qxe6 Rxe6 13.Rxg3 e4
14.Rc3 (5.952.668.958) 505

10/62 196:39 +2.23 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Nc5 5.R4g2 Qf3 6.Ka2 Qxd5
7.h4 Kxg7 8.Ka3 Nd3 9.Qg4 Qc5+
10.Ka2 Nb4+ 11.Kb1 (5.963.859.302) 505

11/62 197:23 +1.62 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Nc5 5.R4g2 Qf3 6.Rg3 Qxd5
7.h4 Nd3 8.Qb6 e4 9.Kb1 Kxg7 10.R3g2 Rc6
11.Qb8 (5.990.337.466) 505

12/63 204:28 +2.44 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Nc5 5.R4g2 Qf3 6.Ka2 Qxd5
7.Rc2 Kxg7 8.Kb1 Nde4 9.Qc4 Nd2+
10.Rxd2 Qxd2 11.h4 e4 12.Ka1 (6.257.243.074) 510

13/65 213:36 +2.51 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Nc5 5.R4g2 Qf3 6.Ka2 Qxd5
7.Rc2 Kxg7 8.Kb1 Na6 9.Qe1 Rb8
10.Rb2 Rxb3 11.Rxb3 Qxb3+ 12.Kc1 e4
13.Rg3 Qxa4 14.Qc3+ Kg8 (6.585.337.268) 513

14/65 216:05 +2.54 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Nc5 5.R4g2 Qf3 6.Ka2 Qxd5
7.Rc2 Kxg7 8.Kb1 Na6 9.Qe1 Rc5
10.Rxc5 Nxc5 11.a5 Qxb3+ 12.Ka1 Qa3+
13.Kb1 (6.667.641.923) 514

15/69 270:57 +2.60 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Re1 Qf3 5.Rg3 Qxd5 6.h4 e4 7.Rg2 Nc5
8.Rb2 Qe5 9.Ka2 Nd3 10.Qd2 Nxe1
11.Qxe1 Rc3 12.Kb1 Kxg7 13.a5 Qd4
14.Ka2 Kh8 (8.550.299.558) 525

16/77 350:11 +2.74 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Re1 Qf3 5.Rg3 Qxd5 6.Rc3 Rb8 7.Qg4 Nc5
8.Qd1 Nxb3+ 9.Rxb3 Qxb3 10.Qxb3 Rxb3
11.Ka2 Rb4 12.Rxe5 Rxa4+ 13.Kb3 Rd4
14.Re7 Kxg7 (11.319.788.596) 538

best move: Rc2xe2 time: 426:18.500 min n/s: 555.040 CPU 100.0% n/s(1CPU): 555.040 nodes: 14.196.960.000

That was build 379. But in the latest build the PV is different and does not go up quite as fast:


[D]2r3k1/2qn1pPp/3n4/3Pp1P1/PQ4R1/1P6/2r1N2P/1K1R1B2 b - -

Engine: Blueberry Beta 4 DM70 Build 392 (Athlon 2009 MHz, 256 MB)
by F. Letouzey, T. Gaksch, E. de Groot

2/50 0:27 -0.48 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rgg1 Rc2 5.Qa3 Qf3 6.Rgf1 Qc3+
7.Kb1 Kxg7 (14.454.970) 524

2/56 0:35 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Ra2+ 6.Kb1 Rb2+
7.Kc1 Rc2+ 8.Kb1 Rb2+ (19.171.430) 535

2/56 0:43 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rb8 5.Qxd6 Qd3 6.Rb1 Qc3+
7.Ka2 Kxg7 8.Qxd7 Kg8 9.Ka3 Qc5+
10.Ka2 Qc2+ 11.Ka1 Qc3+ 12.Ka2 Qc2+ (23.721.914) 541

3/56 0:53 -1.37 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Nc5 (28.890.011) 538

3/56 0:57 -1.19 1...Nb6 2.Qa3 Kxg7 3.Ka1 Ra8 4.h4 Nb5
5.Qb4 Ra2+ 6.Kxa2 Qc2+ 7.Ka1 Qxd1+
8.Kb2 Nxd5 9.Qxb5 Qxf1 10.Qd3 (30.782.640) 532

3/56 1:24 -0.76 1...Kxg7 2.Qa3 Kh8 3.Ka1 f6 4.b4 Rg8
5.g6 Rxg6 6.Rxg6 hxg6 7.a5 Nc4 8.Qa4 (44.467.856) 525

3/56 1:53 -0.01 1...Kxg7 2.Ka1 Nf5 3.Rc4 Rxc4 4.bxc4 Qxc4
5.Qxc4 Rxc4 6.Bh3 Rxa4+ 7.Kb1 Kg6
8.Ng3 Rb4+ 9.Ka2 Ne3 10.Ka3 (58.551.128) 517

3/56 2:50 -0.01 1...Kxg7 2.Ka1 Ra2+ 3.Kxa2 Qc2+
4.Ka3 Qxd1 5.Qxd6 Qa1+ 6.Kb4 Qxf1
7.Qxd7 Qe1+ 8.Kb5 Qxe2+ 9.Rc4 Ra8
10.Kc6 e4 11.Kb5 Rb8+ 12.Ka5 Rxb3 (90.007.078) 527

4/56 3:05 -1.24 1...Kxg7 2.Ka1 Qc5 3.Qxc5 Nxc5 4.Rb4 Nce4 (97.257.131) 525

4/56 3:15 -1.19 1...Nb6 2.Qa3 Kxg7 3.Ka1 Ra8 4.h4 Nb5
5.Qb4 Ra2+ 6.Kxa2 Qc2+ 7.Ka1 Qxd1+
8.Kb2 Nxd5 9.Qxb5 Qxf1 10.Qd3 (102.452.493) 525

4/56 3:29 -1.19 1...Nb6 2.Qa3 Kxg7 3.Ka1 Ra8 4.h4 Nb5
5.Qb4 Ra2+ 6.Kxa2 Qc2+ 7.Ka1 Qxd1+
8.Kb2 Nxa4+ 9.bxa4 Qxf1 10.axb5 Qxe2+
11.Kb1 Qa2+ 12.Kc1 (110.096.420) 526

4/58 6:56 -1.19 1...Nb6 2.Qa3 Rb2+ 3.Qxb2 Ndc4
4.bxc4 Nxc4 5.d6 Nxd6 6.Ka1 Nc4
7.Bh3 f5 8.gxf6 Nxb2 9.f7+ Kxf7
10.g8Q+ Ke7 11.Kxb2 Qc2+ 12.Ka3 Qc5+
13.Ka2 Rxg8 (239.277.984) 574

4/58 6:58 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rgg1 Rc2 5.Qa3 Qf3 6.Qb4 Qf2 7.Qa3 Qf3 (240.459.196) 574

4/58 7:07 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rgg1 Rc2 5.Qa3 Qf3 6.Kb1 Nc4 7.Qb4 Ndb6
8.Kxc2 Qe2+ 9.Kc3 Qf3+ 10.Kc2 (244.536.567) 572

4/58 7:33 0.00 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.R4g2 Rc1+ 6.Rxc1 Qxg2
7.Qxd6 Qxg5 8.Qxd7 Qxc1+ 9.Ka2 Qd2+
10.Ka1 Kxg7 11.Qg4+ Kh6 12.b4 Qxh2
13.Kb1 Qh1+ 14.Kb2 Qh2+ (258.118.306) 569

5/58 8:13 -0.17 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Rdg1 Rc2 5.Qa3 Qxh2 6.Rd1 Kxg7 (278.978.373) 565

5/58 9:31 -0.01 1...Kxg7 2.Ka1 Nf5 3.Rc4 Rxc4 4.bxc4 Qxc4
5.Qxc4 Rxc4 6.Bh3 Rxa4+ 7.Kb1 Kg6
8.Ng3 Rb4+ 9.Ka2 Ne3 10.Ka3 (317.913.655) 556

5/58 9:48 -0.01 1...Kxg7 2.Ka1 Nf5 3.Rc4 Ne3 4.Rxc7 R8xc7
5.Qa3 R2c5 6.Kb1 Rb7 7.d6 e4 8.Ka2 Rc2+
9.Ka1 (325.719.801) 553

5/58 10:58 -0.01 1...Kxg7 2.Ka1 Nf5 3.Rc4 Rxc4 4.bxc4 Qxc4
5.Qxc4 Rxc4 6.Bh3 Rxa4+ 7.Kb1 Kg6
8.Ng3 Rb4+ 9.Ka2 Ne3 10.Ka3 (363.087.735) 551

5/58 12:37 0.00 1...e4 2.Rg3 Rxe2 3.Bxe2 Qc2+ 4.Ka1 Qxe2
5.Rdg1 Rc2 6.R3g2 Ra2+ 7.Kb1 Rb2+
8.Ka1 Ra2+ (414.134.826) 546

5/58 12:42 0.00 1...e4 2.Rg3 Rxe2 3.Bxe2 Qc2+ 4.Ka1 Qxe2
5.Rdg1 Rc2 6.R3g2 Ra2+ 7.Kb1 Rb2+
8.Ka1 Ra2+ (416.977.233) 546

5/58 13:00 0.00 1...e4 2.Rg3 Rxe2 3.Bxe2 Qc2+ 4.Ka1 Qxe2
5.Qd2 Rc2 6.Qxe2 Rxe2 7.b4 Kxg7
8.Kb1 Nc4 9.Kc1 Rxh2 10.a5 Rb2
11.Re1 e3 12.Rexe3 Rxb4 13.a6 Nxe3 (426.052.068) 545

6/58 13:21 -0.25 1...e4 2.Rg3 Rxe2 3.Bxe2 Qc2+ 4.Ka1 Qxe2
5.Qd2 Qxd2 6.Rxd2 Kxg7 7.a5 Re8
8.Kb1 (437.035.936) 545

6/58 15:35 0.00 1...Kxg7 2.Ka1 Nf5 3.Rc4 Rxc4 4.bxc4 Qxc4
5.Qxc4 Rxc4 6.Bh3 Rxa4+ 7.Kb1 Kg6
8.Ng3 Rb4+ 9.Ka2 Ra4+ 10.Kb1 (500.067.616) 534

6/58 17:09 0.00 1...Kxg7 2.Ka1 Nf5 3.Rc4 Rxc4 4.bxc4 Qxc4
5.Qxc4 Rxc4 6.Bh3 Rxa4+ 7.Kb1 Rb4+
8.Kc2 Ne3+ 9.Kc3 Rb8 10.Rd3 Rc8+
11.Kd2 Nc4+ 12.Ke1 e4 (547.063.304) 531

6/58 20:29 0.00 1...Kxg7 2.Ka1 Nf5 3.Rc4 Rxc4 4.bxc4 Qxc4
5.Qxc4 Rxc4 6.Bh3 Rxa4+ 7.Kb1 Rb4+
8.Ka1 Ne3 9.Rd3 Ra4+ 10.Kb1 Ng4
11.Bxg4 Rb4+ 12.Ka2 Rxg4 13.Rg3 Ra4+
14.Kb3 Nc5+ (653.359.632) 531

6/60 21:26 +0.48 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qg3 7.Rhc1 Qf2
8.a5 Kxg7 9.Rh1 Nde4 10.Rhg1 Ng3 (685.311.711) 532

6/60 22:14 +0.48 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qf2 7.Rh6 Nxb3+
8.Qxb3 Qd4+ 9.Ka2 Rc4 10.Ka3 Qc5+
11.Ka2 f5 12.gxf6 Rc2+ 13.Ka1 Qd4+
14.Rb2 Qxb2+ (713.891.119) 535

6/60 25:25 +0.48 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rg3 Qxe2
5.Qd2 Qxd2 6.Rxd2 Nce4 7.Ra2 Nxg3
8.hxg3 Nb7 9.a5 Ra8 10.Kb1 Nxa5
11.b4 (844.740.213) 553

7/60 25:41 +0.19 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qg3 7.Rhc1 Qf2
8.Rd1 Kxg7 (854.072.170) 554

8/60 30:36 +0.19 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qg3 7.Rhc1 Qf2
8.Rd1 Kxg7 (1.014.313.598) 552

9/60 99:02 +0.65 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qg2 7.Rhd1 Kxg7
8.Qc3 Nd7 9.Qd2 Qxd2 10.Rxd2 Nc5
11.Ka2 (3.176.905.153) 534

10/60 104:28 +0.74 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qg2 7.Rhd1 Kxg7
8.Qc3 Nd7 9.Qd2 Rc2 10.Qxg2 Rxg2 (3.378.100.256) 538

11/62 168:32 +0.97 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qc2 7.Rhc1 Qf2
8.Rf1 Qg2 9.Rfd1 Kxg7 10.Qc3 Nd3
11.Rd2 Rxc3 12.Rxg2 e4 13.a5 Nb4 (5.462.209.844) 540

12/62 198:50 +0.77 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qc2 7.Rhc1 Qf2
8.Rd1 Kxg7 9.Rf1 Qc2 10.Rfd1 Ra8 (6.476.945.125) 542

13/66 281:38 +0.77 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qc2 7.Rhc1 Qf2
8.Rd1 Kxg7 9.Rf1 Qc2 10.Rfd1 Ra8 (9.275.792.620) 548

14/67 351:11 +0.84 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qg3 7.Ka2 Nce4
8.Ka1 Rc2 9.Rbc1 Rxc1+ 10.Rxc1 Qxg5
11.Rc7 Qg4 12.Qb6 Kxg7 (11.576.698.567) 549

15/67 374:32 +0.97 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Nc5 4.Rb1 Qxe2
5.Rgg1 Qxh2 6.Rh1 Qc2 7.Rhc1 Qf2
8.a5 Qe3 9.Rg1 Nf5 10.Rg2 Qf3
11.Rgb2 Qxd5 12.Rf1 e4 13.Qb6 (12.304.026.000) 547

Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
User avatar
Eelco de Groot
Posts: 4564
Joined: Sun Mar 12, 2006 2:40 am
Full name:   

Re: Blueberry Beta 4 Build 392 result arasan 11.201

Post by Eelco de Groot »

Finally after more than 800 minutes iteration 16:

16/76 838:11 +2.63 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Re1 Qf3 5.Rg3 Qxd5 6.Rc3 Nc5 7.Qa3 Nde4
8.Rxe4 Qxe4 9.Ka2 Qd5 10.a5 Kxg7
11.Kb1 Qd4 12.Rc4 Qd3+ 13.Ka2 Qe2+
14.Kb1 Qf1+ (27.665.075.443) 550

17/76 850:21 +2.63 1...Rxe2 2.Bxe2 Qc2+ 3.Ka1 Qxe2
4.Re1 Qf3 5.Rg3 Qxd5 6.Rc3 Nc5 7.Qa3 Nde4
8.Rxe4 Qxe4 9.Ka2 Qd5 10.a5 Kxg7
11.Kb1 Qd4 12.Rc4 Qd3+ 13.Ka2 Qe2+
14.Kb1 Qf1+ (28.127.375.501) 551


Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
Top
Howard E
Posts: 261
Joined: Wed Mar 08, 2006 8:49 pm

Re: new test positions

Post by Howard E »

[D] 2r3k1/2qn1pPp/3n4/3Pp1P1/PQ4R1/1P6/2r1N2P/1K1R1B2 b - a3 0 1

ProDeo Ares: i7920, vista premium 64bit, 512 hash

12 00:04 10.962.903 2.192.580 +0.61 Rxe2 Bxe2 Qc2+ Ka1 Nc5 Rg3 Qxe2 Qd2 Qxd2 Rxd2 Nde4 Rdg2 Nxg3 Rxg3 Kxg7 Rc3 f5 gxf6+ Kxf6
12 00:08 18.329.896 2.221.805 +0.61 Rb2+
12 00:09 20.475.422 2.213.559 +1.32 Rb2+ Kxb2 Qc2+ Ka3 Qxd1 Qxd6 Rb8 Qxb8+ Nxb8 Rg1 Qxd5 Rg4 Qc5+ Rb4 Kxg7
13 00:09 22.604.393 2.511.599 +1.32 Rb2+ Kxb2 Qc2+ Ka3 Qxd1 Qxd6 Rb8 Qxb8+ Nxb8 Rg1 Qa1+ Kb4 Kxg7 Bg2 Qb2 Nc3 Nd7
14 00:20 47.661.174 2.383.058 +1.68 Rb2+ Kxb2 Qc2+ Ka3 Qxd1 Qxd6 Rb8 b4 Rc8 Qc6 Rxc6 dxc6 Nb6 a5 Qa1+ Kb3 Nd5 Kc4 Ne3+ Kb5 Qxf1 Re4 Qxe2+ Kb6
15 00:42 101.964.004 2.371.255 +2.32 Rb2+ Kxb2 Qc2+ Ka3 Qxd1 Qxd6 Rb8 b4 Rc8 Qc6 Rxc6 dxc6 Qd3+ Ka2 Qc2+ Ka3 Qxc6 Bg2 Qc2 Re4 Qd3+ Ka2 Qd2+ Kb1
16 01:15 183.536.800 2.447.157 +2.32 Rb2+ Kxb2 Qc2+ Ka3 Qxd1 Qxd6 Rb8 b4 Rc8
17 03:50 560.266.755 2.435.942 +2.58 Rb2+ Kxb2 Qc2+ Ka3 Qxd1 Qxd6 Rb8 b4 Rc8 Qc6 Rxc6

sticks to Rb2. If one enters the first few forced moves of this line other programs like it too, even though they initially like Qxd2.

Thanks for the positions.
Top
Howard E
Posts: 261
Joined: Wed Mar 08, 2006 8:49 pm

Re: new test positions

Post by Howard E »

ProDeo Ares: i7920,vista 64bit pemium, 512 hash

[D] 7b/8/kq6/8/8/1N2R3/K2P4/8 w - - 0 1

12 00:02 6.619.089 3.309.544 -5.09 Re8 Qf6 Ra8+ Kb7 Rxh8 Qxh8 d4 Kc7 Kb2 Kd6 Kc2 Qh7+ Kc3 Qe4
12 00:02 8.226.132 4.113.066 -4.35 Kb1 Qg6+ Kc1 Kb5 Kd1 Qg4+ Kc2 Qc4+ Kd1 Qf1+ Re1 Qf3+ Kc2 Kc4 Nc1 Bd4 Re2
13 00:05 14.901.464 2.838.374 -4.14 Kb1 Qg6+ Kc1 Qc6+ Kd1 Qh1+ Kc2 Qc6+ Kd1 Qh1+ Kc2 Qc6+
14 00:11 32.669.061 2.969.914 -4.47 Kb1 Qg6+ Kc1 Kb5 Kd1 Qb1+ Ke2 Bf6 Kf2 Bg5 Rc3
15 00:16 49.345.414 3.084.088 -4.67 Kb1 Qg6+ Kc1 Qg5 Kd1 Qg1+ Kc2 Qg6+ Kc1 Qg5 Kd1 Qg1+ Kc2 Qg6+
15 00:45 135.943.477 3.020.966 -4.67 Nd4
15 00:46 138.294.581 3.006.403 -4.67 Nd4
15 00:46 139.084.923 3.023.585 -0.16 Nd4 Qxd4 Ra3+ Kb7 Rg3 Qxd2+ Kb3 Kc6 Rg6+ Bf6 Rxf6+ Kd5 Ra6 Qd4 Ra5+ Ke6 Rh5 Qe4 Rh6+ Kd5 Kc3 Ke5
16 00:54 166.767.737 3.088.291 -0.15 Nd4 Qxd4 Ra3+ Kb7 Rb3+ Kc7 Rc3+ Kd6 Rd3 Bg7 Rxd4+ Kc6 Ka3 Bxd4 Kb4 Kb6 Kb3 Kc5 Kc2 Kd5 Kd3 Bc5 Kc3 Kc6
17 00:55 172.029.331 3.127.806 -0.11 Nd4 Qxd4 Ra3+ Kb7 Rb3+ Kc7 Re3 Qxd2+ Ka3 Qb2+ Ka4 Kd6 Ra3 Qc2+ Rb3 Kc6 Kb4 Qc5+ Ka4 Qc4+
18 00:58 182.651.258 3.149.159 -0.16 Nd4 Qxd4 Ra3+ Kb7 Rb3+ Kc7 Rc3+ Kd6 Rd3 Ke5 Rxd4 Kxd4 Kb3 Be5 Kc2 Bf6 d3 Bg7 Kd2 Be5 Ke2 Bd6 Kd2 Bb4+
19 01:12 229.151.495 3.139.061 -0.11 Nd4 Qxd4 Ra3+ Kb7 Rb3+ Kc6 Re3 Qxd2+ Kb3 Qb2+ Ka4 Be5 Re1
20 01:20 257.810.108 3.222.626 -0.14 Nd4 Qxd4 Ra3+ Kb7 Rb3+ Kc6 Rc3+ Kd5 Rd3 Be5 Rxd4+ Kxd4 Kb3 Bf4 Kc2 Bc7 d3 Kc5 Kc3 Bd6 Kd2 Kd5 Ke3 Be5
21 01:37 325.001.292 3.316.339 -0.08 Nd4 Qxd4 Ra3+ Kb7 Rb3+ Kc7 Rd3 Qb2+
22 02:09 449.049.581 3.454.227 -0.17 Nd4 Qxd4 Ra3+ Kb7 Rb3+ Kc7 Rc3+ Kb6 Rb3+ Kc5 Rc3+ Kb4 Rb3+ Kc4 Rb4+ Kd5 Rxd4+ Kxd4 Kb2 Kd3+ Kb3 Bf6 Kb4 Be5
23 03:41 801.549.498 3.610.583 -0.08 Nd4 Qxd4 Ra3+ Kb7 Rb3+ Kc7 Rc3+ Kd6 Kb3 Qxd2 Rc6+ Kxc6 Ka4 Qf4+ Kb3 Qe3+ Ka4 Qe4+ Kb3 Qb1+ Ka4 Qb5+ Ka3 Qa5+


A fun position . Other programs find it immediately while others are stuck here. Only upon entering Nd4 do they instantly see the draw.
Must be pruned out.
Top

Return to “Computer Chess Club: General Topics”