Bug with MultiPV in TOGA II - Alternatives with MultiPV?

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

Moderator: Ras

rj909

Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by rj909 »

I'm currently using TOGA II to assist in finding tactical problems for my new
tactical problem site Chess Tempo. I currently use MultiPV mode (with MultiPV set to 2) to help eliminate problems which have ambiguous solutions (second best move too close in score to best move). Unfortunately TOGA II is sometimes way out on its scoring with MultiPV. Some positions when analyzed with standard options (i.e. no MultiPV analyze position after each move separately) the score for two alternatives may be very close (say <0.4). However in MultiPV mode the difference is >2.0 . I think this is a straight out bug as giving more time to MultiPV does not help (i.e. its not just that less of search space is explored due to processing the alternatives).

Has anyone else noticed these issues on TOGA II (I've tried both the latest beta and the previous stable releases and they appear to behave in the same manner)?

The following position triggers the problem to me:

6k1/6pp/pp6/8/1P3b2/N1P2PPP/b2r4/1R4K1 b - - 0 1

I've had a quick look at the code and haven't been able to track down the problem yet so I'm considering alternative engines. Can anyone suggest a strong engine with UCI and support for the MultiPV option? As I'm searching for tactics, tactical strength is obviously more important to me than positional sophistication. I'd prefer open source as I may need to make changes tot he engine to help with problem generation, although I'd consider commercial engines if they ran on both linux and Mac OS X.

Thanks,
Richard.
Marc MP

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by Marc MP »

rj909 wrote:I'm currently using TOGA II to assist in finding tactical problems for my new
tactical problem site Chess Tempo. I currently use MultiPV mode (with MultiPV set to 2) to help eliminate problems which have ambiguous solutions (second best move too close in score to best move). Unfortunately TOGA II is sometimes way out on its scoring with MultiPV. Some positions when analyzed with standard options (i.e. no MultiPV analyze position after each move separately) the score for two alternatives may be very close (say <0.4). However in MultiPV mode the difference is >2.0 . I think this is a straight out bug as giving more time to MultiPV does not help (i.e. its not just that less of search space is explored due to processing the alternatives).

Has anyone else noticed these issues on TOGA II (I've tried both the latest beta and the previous stable releases and they appear to behave in the same manner)?

The following position triggers the problem to me:

6k1/6pp/pp6/8/1P3b2/N1P2PPP/b2r4/1R4K1 b - - 0 1

I've had a quick look at the code and haven't been able to track down the problem yet so I'm considering alternative engines. Can anyone suggest a strong engine with UCI and support for the MultiPV option? As I'm searching for tactics, tactical strength is obviously more important to me than positional sophistication. I'd prefer open source as I may need to make changes tot he engine to help with problem generation, although I'd consider commercial engines if they ran on both linux and Mac OS X.

Thanks,
Richard.
Hi Richard,

I would highly recommend you the open-source UCI-program Glaurung 1.2.1 (which runs on both linux and Mac OS X).

For sure Glaurung is one of the best free engine for tactics, support MultiPV and searchmoves uci-command (that enables you to include or exclude moves from analysis - that is what you need for your tactical positions).

http://www.glaurungchess.com/

Also good free uci-compatible tacticians are Rybka 1.0 Beta (set the settings at "very tactical"), but this one doesn't support MultiPV and SlowChess Blitz WV 2.1 (http://www.3dkingdoms.com/chess/slow.htm), this one support MultiPV and can also be used in its own GUI!

Happy solving!

Marc
User avatar
Mike S.
Posts: 1480
Joined: Thu Mar 09, 2006 5:33 am

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by Mike S. »

I tried your position with Toga II 1.2.1a (default settings), up to depth 15 in multi-pv, and up to depth 14 after each of the two moves:

1) Multi-PV (2)

Code: Select all

 15	 1:52 	+10.63 	1...Lxb1 2.gxf4 Ld3 3.h4 Ta2 4.h5 Txa3 5.Kg2 Txc3 6.Kf2 Kf7 7.Ke3 Lb5+ 8.Kd4 Txf3 9.Ke4 Tb3 10.Kf5 Txb4 11.Ke5 (93.152.175) 828.5
 15	 1:54 	+7.42 	1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Sxb1 Txf3 5.h4 Txg3 6.Kd1 Tg2 7.h5 Lf4 8.Sa3 Td2+ 9.Ke1 Td3 10.Sc4 b5 11.Sb6 Txc3 12.Sd5 (94.787.045) 828.5
(L=Bishop)

2) singlemode after 1...Bxb1

Code: Select all

 14.01	 0:04 	-10.89 	2.Sxb1 Td1+ 3.Kf2 Txb1 4.Ke2 Tb2+ 5.Kd1 Lxg3 6.Kc1 Tf2 7.Kd1 Txf3 8.c4 Ld6 9.b5 a5 10.h4 Tc3 11.Kd2 Txc4 (4.250.620) 870.0
Here, it switched from 2.gxf4 to 2.Nxb1 for the 2nd ply but the eval is very similar.

3) singlemode after 1...Be3+

Code: Select all

 14.01	 0:03 	-7.34 	2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Sxb1 Txf3 5.h4 Txg3 6.Ke2 Lf4 7.Kd1 Th3 8.Kc2 Txh4 9.c4 Le5 10.Sd2 g5 11.Kd3 Td4+ 12.Ke3 (3.279.680) 903.3
Looks ok to me. (Maybe the small differences were caused by remaining hashtable entries.)

I think it should almost always look like above (or preferable, entirely identical) if the comparison is done like

1) Multi-PV at depth x, to
2) singlemode at depth x-1 after the first moves each.

One ply +/- can of course make a big difference. :mrgreen: Also, sometimes there are huge differences betwenn multi-pv and singlemode, but that should be rare cases where it's caused by other effects.
Regards, Mike
User avatar
Eelco de Groot
Posts: 4660
Joined: Sun Mar 12, 2006 2:40 am
Full name:   Eelco de Groot

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by Eelco de Groot »

Could it maybe be an interface problem Richard? I tried it not with default Toga but with some Toga II 1.3x4 settings, but that should not make much difference. I could not see much out of the ordinary. Toga does not support Exclude Move so I could only compare the searches going one ply deeper but I saw no big jumps in evaluation either way, going one move deeper or with Two best moves;

Output in Shredder 9.1 UCI GUI from Toga 1.3 Checkov:

8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

6k1/6pp/pp6/8/1P3b2/N1P2PPP/b2r4/1R4K1 b - -

Engine: Toga 1.3 "Checkov" (Athlon 2009 MHz, Windows XP, 256 MB)
gemaakt door Thomas Gaksch and Fabien Letouzey

10/22 0:00 +6.26 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1
4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+
8.Kf2 (143.453)

10/22 0:00 +6.78 1...Lxb1 2.Pxb1 Td1+ 3.Kf2 Lxg3+
4.Kxg3 Txb1 5.Kf4 Tb3 6.h4 Txc3
7.Ke4 Tc4+ 8.Ke5 Txb4 (257.180)

11/23 0:00 +6.88 1...Lxb1 2.Pxb1 Td1+ 3.Kf2 Lxg3+
4.Kxg3 Txb1 5.Kf4 Tb3 6.h4 Txc3
7.Kg5 Tc4 8.f4 Txb4 (414.368)

12/44 0:02 +8.48++ 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5
5.Pc4 Lxc4 6.bxc5 Ld5 7.f5 Lxf3 8.h4 (2.455.845) 1030

13/44 0:02 +8.59 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5
5.bxc5 Txa3 6.Kf2 Lf5 7.Ke1 Tc3
8.Kd2 Txc5 (2.574.769) 1020

14/44 0:02 +8.76 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5
5.bxc5 Txa3 6.Kf2 Lf5 7.h4 Tc3 8.Ke2 a5
9.Kd2 Txc5 (2.942.465) 1009

15/44 0:04 +8.93 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5
5.bxc5 Txa3 6.Kf2 Lf5 7.Ke2 Tc3 8.h4 a5
9.Kd2 Txc5 10.Ke3 Tc3+ 11.Kd4 Txf3 (4.187.883) 980

16/44 0:07 +9.22 1...Lxb1 2.gxf4 Ld3 3.b5 Ta2 4.bxa6 Lxa6
5.Pb1 Ld3 6.h4 Lxb1 7.Kf1 Ld3+ 8.Ke1 Kf7
9.h5 Ke7 10.h6 g6 11.f5 Lxf5 (7.339.169) 961

17/44 0:11 +9.34 1...Lxb1 2.gxf4 Ld3 3.b5 Ta2 4.bxa6 Lxa6
5.Pb1 Ld3 6.h4 Lxb1 7.Kf1 Ld3+ 8.Ke1 Kf7
9.h5 Ke7 10.f5 Lxf5 11.c4 Kd6 (10.721.073) 946

18/67 1:41 +10.33++ 1...Lxb1 2.gxf4 Ld3 3.h4 Ta2 4.f5 Txa3
5.Kf2 Txc3 6.f6 gxf6 7.h5 Kf7 8.Ke1 Lf5
9.Ke2 Tb3 10.b5 Txb5 11.Kf2 (86.224.589) 852

19/67 2:04 +9.93-- 1...Lxb1 2.gxf4 Ld3 3.h4 Ta2 4.f5 Txa3
5.Kf2 Ta2+ 6.Ke3 a5 7.Kxd3 Td2+ (106.335.315) 857

20/68 6:46 +12.58++ 1...Lxb1 2.gxf4 Ld3 3.f5 Ta2 4.f4 Txa3
5.Kf2 Lxf5 6.Ke3 Txc3+ 7.Kd4 Txh3
8.Ke5 a5 9.bxa5 bxa5 10.Kxf5 a4
11.Kg4 a3 12.Kxh3 a2 13.Kg4 Kf7
14.Kf5 Ke7 (333.306.532) 820

21/68 7:46 +12.18-- 1...Lxb1 2.gxf4 Ld3 3.f5 Td1+ 4.Kg2 Lf1+
5.Kg1 Lxh3+ (382.490.516) 820

Two best moves are 1...Bxb1 and 1...Be3+

8 0:00 +6.95 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Ta2+ 7.Ke3 (39.564)
7 0:00 +6.07 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Th1+ 7.Ke2 (28.170)
-----
8 0:00 +6.95 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Ta2+ 7.Ke3 (39.564)
8 0:00 +6.28 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Th2 (40.521)
-----
9 0:00 +7.04 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Kf7 (62.644)
8 0:00 +6.28 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Th2 (40.521)
-----
9 0:00 +7.04 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Kf7 (62.644)
9 0:00 +6.26 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 (64.162)
-----
10 0:00 +7.26 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Kf7 7.Kg3 (136.907)
9 0:00 +6.26 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 (64.162)
-----
10 0:00 +7.26 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Kf7 7.Kg3 (136.907)
10 0:00 +6.26 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 (176.825)
-----
11 0:00 +7.44 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 a5 7.Ke3 Lf5+ 8.Kd4 Td3+ 9.Kc4 Txf3 (257.046)
10 0:00 +6.26 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 (176.825)
-----
11 0:00 +7.44 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 a5 7.Ke3 Lf5+ 8.Kd4 Td3+ 9.Kc4 Txf3 (257.046)
11 0:00 +6.74 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Td3 8.Pf1 (302.709)
-----
12 0:00 +8.48 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.Pc4 Lxc4 6.bxc5 Ld5 7.f5 Lxf3 8.h4 (542.455)
11 0:00 +6.74 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Td3 8.Pf1 (302.709)
-----
12 0:00 +8.48 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.Pc4 Lxc4 6.bxc5 Ld5 7.f5 Lxf3 8.h4 (542.455)
12 0:00 +6.77 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Ke2 Lf4 7.Kf2 Td3 8.c4 Kf7 (656.352)
-----
13 0:00 +8.59 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Lf5 7.Ke1 Tc3 8.Kd2 Txc5 (903.179)
12 0:00 +6.77 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Ke2 Lf4 7.Kf2 Td3 8.c4 Kf7 (656.352)
-----
13 0:00 +8.59 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Lf5 7.Ke1 Tc3 8.Kd2 Txc5 (903.179) 960
13 0:01 +6.78 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 Tc3 9.Ke1 Lxd2+ 10.Kxd2 Txc4 (1.274.697) 960
-----
14 0:01 +8.76 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Lf5 7.h4 Tc3 8.Ke2 a5 9.Kd2 Txc5 (1.532.752) 948
13 0:01 +6.78 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 Tc3 9.Ke1 Lxd2+ 10.Kxd2 Txc4 (1.274.697) 948
-----
14 0:01 +8.76 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 Lf5 7.h4 Tc3 8.Ke2 a5 9.Kd2 Txc5 (1.532.752) 955
14 0:02 +7.12 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Kd1 Th2 7.c4 Tb2 8.Pc3 Txb4 9.Pd5 Tb1+ 10.Ke2 (2.461.004) 955
-----
15 0:06 +9.39 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 a5 7.Ke1 Ta2 8.Kd1 Tc2 9.f5 Lxf5 10.f4 Txc5 (5.585.006) 928
14 0:02 +7.12 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Kd1 Th2 7.c4 Tb2 8.Pc3 Txb4 9.Pd5 Tb1+ 10.Ke2 (2.461.004) 928
-----
15 0:06 +9.39 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 bxc5 5.bxc5 Txa3 6.Kf2 a5 7.Ke1 Ta2 8.Kd1 Tc2 9.f5 Lxf5 10.f4 Txc5 (5.585.006) 927
15 0:07 +6.91 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 Tc3 9.Pe4 Txc4 10.Kf3 Le5 (7.249.054) 927
-----
16 0:12 +9.66 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.Pb5 axb5 5.cxb5 Kf7 6.f5 Kf6 7.h4 Kxf5 8.Kh1 Kf4 9.Kg1 Kxf3 (11.669.955) 924
15 0:07 +6.91 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 Tc3 9.Pe4 Txc4 10.Kf3 Le5 (7.249.054) 924
-----
16 0:12 +9.66 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.Pb5 axb5 5.cxb5 Kf7 6.f5 Kf6 7.h4 Kxf5 8.Kh1 Kf4 9.Kg1 Kxf3 (11.669.955) 922
16 0:14 +7.17 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Tg2 7.c4 Tb2 8.Pc3 Txb4 9.Pd5 Tb1+ 10.Ke2 Ld4 11.h5 (13.831.711) 922
-----
17 0:37 +10.16 1...Lxb1 2.gxf4 Ld3 3.h4 Ta2 4.h5 Txa3 5.Kf2 Txc3 6.Ke1 Lb5 7.Kf2 a5 8.bxa5 bxa5 9.Ke1 Txf3 10.f5 a4 11.Kd2 Txf5 (32.969.417) 876
16 0:14 +7.17 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Tg2 7.c4 Tb2 8.Pc3 Txb4 9.Pd5 Tb1+ 10.Ke2 Ld4 11.h5 (13.831.711) 876
-----
17 0:37 +10.16 1...Lxb1 2.gxf4 Ld3 3.h4 Ta2 4.h5 Txa3 5.Kf2 Txc3 6.Ke1 Lb5 7.Kf2 a5 8.bxa5 bxa5 9.Ke1 Txf3 10.f5 a4 11.Kd2 Txf5 (32.969.417) 880
17 0:41 +7.15 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Tg2 7.c4 Tb2 8.Pc3 Txb4 9.Pd5 Tb1+ 10.Kc2 Te1 11.Kd3 Lc5 (36.641.586) 880
-----
18 1:05 +10.47 1...Lxb1 2.gxf4 Ld3 3.h4 Ta2 4.f5 Txa3 5.Kf2 Txc3 6.f6 gxf6 7.h5 Tb3 8.Ke1 Txb4 9.Kd2 Lb5 10.Ke3 Tb3+ 11.Ke4 Lc6+ 12.Kd4 Lxf3 (57.552.009) 873
17 0:41 +7.15 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Tg2 7.c4 Tb2 8.Pc3 Txb4 9.Pd5 Tb1+ 10.Kc2 Te1 11.Kd3 Lc5 (36.641.586) 873
-----
18 1:05 +10.47 1...Lxb1 2.gxf4 Ld3 3.h4 Ta2 4.f5 Txa3 5.Kf2 Txc3 6.f6 gxf6 7.h5 Tb3 8.Ke1 Txb4 9.Kd2 Lb5 10.Ke3 Tb3+ 11.Ke4 Lc6+ 12.Kd4 Lxf3 (57.552.009) 879
18 1:15 +7.41 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Kd1 Tg3 7.Kc2 Txg4 8.Kd3 Lf4 9.Pa3 Tg3+ 10.Kd4 Ld2 11.Pb1 Le3+ 12.Ke5 Tf3 (65.988.040) 879
-----
19 2:46 +11.48 1...Lxb1 2.gxf4 Ld3 3.f5 Ta2 4.f6 Txa3 5.Kf2 gxf6 6.Ke3 Txc3 7.h4 Tb3 8.f4 Lf5+ 9.Ke2 Txb4 10.h5 a5 11.Kd2 b5 12.Kc3 Txf4 13.Kb2 Tb4+ 14.Ka3 (142.716.095) 856
18 1:15 +7.41 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.g4 Txh3 6.Kd1 Tg3 7.Kc2 Txg4 8.Kd3 Lf4 9.Pa3 Tg3+ 10.Kd4 Ld2 11.Pb1 Le3+ 12.Ke5 Tf3 (65.988.040) 856
-----
19 2:46 +11.48 1...Lxb1 2.gxf4 Ld3 3.f5 Ta2 4.f6 Txa3 5.Kf2 gxf6 6.Ke3 Txc3 7.h4 Tb3 8.f4 Lf5+ 9.Ke2 Txb4 10.h5 a5 11.Kd2 b5 12.Kc3 Txf4 13.Kb2 Tb4+ 14.Ka3 (142.716.095) 864
19 3:11 +7.57 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Lf4 7.Kc2 Th3 8.Kb3 Kf7 9.Pa3 Le5 10.Pb1 b5 11.h5 Ke6 12.Kc2 Txh5 (165.702.108) 864
-----
20 12:56 +13.81 1...Lxb1 2.gxf4 Ld3 3.f5 Ta2 4.f4 Txa3 5.Kf2 Txc3 6.Ke3 Lxf5+ 7.Kd4 Txh3 8.Kd5 h5 9.Ke5 h4 10.Kxf5 Tb3 11.b5 h3 12.bxa6 Tb5+ 13.Ke4 h2 14.Ke3 h1D (626.146.044) 806
19 3:11 +7.57 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Lf4 7.Kc2 Th3 8.Kb3 Kf7 9.Pa3 Le5 10.Pb1 b5 11.h5 Ke6 12.Kc2 Txh5 (165.702.108) 806
-----
20 12:56 +13.81 1...Lxb1 2.gxf4 Ld3 3.f5 Ta2 4.f4 Txa3 5.Kf2 Txc3 6.Ke3 Lxf5+ 7.Kd4 Txh3 8.Kd5 h5 9.Ke5 h4 10.Kxf5 Tb3 11.b5 h3 12.bxa6 Tb5+ 13.Ke4 h2 14.Ke3 h1D (626.146.044) 810
20 13:26 +7.57 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Lf4 7.Kc2 Th3 8.Kb3 Kf7 9.Pa3 Le5 10.Pb1 b5 11.h5 Ke6 12.Kc2 Txh5 (653.457.633) 810
-----
21 36:40 +15.82 1...Lxb1 2.gxf4 Ld3 3.f5 Ta2 4.h4 Txa3 5.Kf2 Txc3 6.h5 Lxf5 7.Ke2 a5 8.bxa5 bxa5 9.Kd2 Tc2+ 10.Ke3 a4 11.Kf4 a3 12.Kxf5 a2 13.f4 Tc5+ 14.Ke4 Txh5 (1.778.732.045) 808
20 13:26 +7.57 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Lf4 7.Kc2 Th3 8.Kb3 Kf7 9.Pa3 Le5 10.Pb1 b5 11.h5 Ke6 12.Kc2 Txh5 (653.457.633) 808
-----
21 36:40 +15.82 1...Lxb1 2.gxf4 Ld3 3.f5 Ta2 4.h4 Txa3 5.Kf2 Txc3 6.h5 Lxf5 7.Ke2 a5 8.bxa5 bxa5 9.Kd2 Tc2+ 10.Ke3 a4 11.Kf4 a3 12.Kxf5 a2 13.f4 Tc5+ 14.Ke4 Txh5 (1.778.732.045) 815
21 39:12 +7.54 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.h4 Txg3 6.Kd1 Lf4 7.Kc2 Th3 8.Pa3 Le5 9.Pb1 Kf7 10.h5 a5 11.Kb3 axb4 12.Kxb4 Txh5 13.Pd2 Ld6+ 14.Kc4 Tc5+ (1.918.422.694) 815


After 1. ... Be3+


6k1/6pp/pp6/8/1P6/N1P1bPPP/b2r4/1R4K1 w - -

Engine: Toga 1.3 "Checkov" (256 MB)
gemaakt door Thomas Gaksch and Fabien Letouzey

10/25 0:00 -6.26 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 (101.562)

11/27 0:00 -6.74 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Pd2 Lf4 7.c4 Td3 8.Pf1 (165.095)

12/27 0:00 -6.77 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Ke2 Lf4 7.Kf2 Td3 8.c4 Kf7 (246.257)

13/35 0:01 -6.78 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Pd2 Lf4 7.c4 Te3+ 8.Kf2 Tc3
9.Ke1 Lxd2+ 10.Kxd2 Txc4 (1.254.536) 1000

14/37 0:01 -7.12 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Kd1 Th2 7.c4 Tb2 8.Pc3 Txb4
9.Pd5 Tb1+ 10.Ke2 (1.922.019) 996

15/39 0:03 -7.12 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Pd2 Lxd2+ 7.Kxd2 Tg3
8.c4 Txg4 9.Kd3 Kf7 10.Kc3 Te4
11.Kd3 (3.662.411) 993

16/42 0:07 -7.04 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.h4 Txg3 6.Pd2 Lf4 7.Pc4 b5 8.Pb6 Txc3
9.Pd5 Tc4 10.Kf2 Kf7 (7.268.887) 962

17/45 0:14 -7.35 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.c4 Txg3 6.Pc3 Txh3 7.Pd5 Ld4 8.Kd2 g5
9.Pe7+ Kf7 10.Pc6 Lc3+ 11.Ke2 h5
12.c5 bxc5 13.bxc5 g4 14.Pd8+ Ke7
15.Pc6+ (13.643.251) 956

18/45 0:22 -7.35 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.Ke2 Txg3 6.Pa3 Txh3 7.Pc4 Lg5
8.Pxb6 Te3+ 9.Kf2 Txc3 10.Ke2 Kf7
11.Pd5 Tc8 12.Pe3 (21.566.702) 950

19/49 0:39 -7.51 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Kd1 Tg3 7.Kc2 Txg4 8.Pa3 Lg5
9.Kb3 Lf6 10.Pc4 Tg3 11.Pxb6 Txc3+
12.Ka4 Kf7 (37.479.627) 946

20/51 1:11 -7.48 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Kd1 Tg3 7.Kc2 Txg4 8.Pa3 Lg5
9.Kb3 Lf6 10.Pc4 Tg3 11.Pxb6 Txc3+
12.Ka4 Kf7 13.Pd5 (67.044.359) 940

21/54 2:07 -7.66 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Kd1 Tg3 7.Kc2 Txg4 8.Kd3 Lf4
9.Pa3 Tg3+ 10.Kd4 Lg5 11.Pc4 Lf6+
12.Ke4 Lxc3 13.Pxb6 Tg4+ 14.Kd3 Lxb4 (119.528.428) 934

22/54 4:02 -7.85 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Kd1 Tg3 7.Kc2 Txg4 8.Pa3 Lg5
9.Kd3 Tg3+ 10.Kc4 Lf6 11.Pb1 b5+
12.Kb3 Kf7 13.Kc2 Tf3 14.Kb3 h5
15.Kc2 (223.937.857) 924

Eelco
Last edited by Eelco de Groot on Thu May 03, 2007 4:04 am, edited 1 time in total.
rj909

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by rj909 »

Hi thanks for looking at this, the problem I have is that I get the following without multipv:
info multipv 1 depth 15 seldepth 47 score cp 780 lowerbound time 18503 nodes 14512450 pv a2b1 a3b1 d2d1 g1f2 d1b1 g3f4 a6a5 b4a5 b6a5 f2e2 a5a4 c3c4 a4a3 c4c5
info depth 15 seldepth 47 time 18503 nodes 14512450 nps 784339

(and I find that once I've used the multipv mode I need to restart the engine to return to default behaviour as setting multipv back to 1 doesn't return to default behaviour [for me at least]).

So the score suddenly jumps from 7.80 without multipv to 10.63 with multipv.

Maybe I'm doing something very silly :-)

This is Toga II 1.2.1a built under Mac OS X (intel).

Regards,
Richard.
User avatar
Eelco de Groot
Posts: 4660
Joined: Sun Mar 12, 2006 2:40 am
Full name:   Eelco de Groot

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by Eelco de Groot »

Eelco de Groot wrote:

After 1. ... Be3+


6k1/6pp/pp6/8/1P6/N1P1bPPP/b2r4/1R4K1 w - -

Engine: Toga 1.3 "Checkov" (256 MB)
gemaakt door Thomas Gaksch and Fabien Letouzey

.
.
.

22/54 4:02 -7.85 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Kd1 Tg3 7.Kc2 Txg4 8.Pa3 Lg5
9.Kd3 Tg3+ 10.Kc4 Lf6 11.Pb1 b5+
12.Kb3 Kf7 13.Kc2 Tf3 14.Kb3 h5
15.Kc2 (223.937.857) 924

Eelco
At 23 plies deep the score goes down again but only after a deep search of 78 minutes after entering 1. ... Be3+


8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

6k1/6pp/pp6/8/1P6/N1P1bPPP/b2r4/1R4K1 w - -


22/54 4:02 -7.85 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3
5.g4 Txh3 6.Kd1 Tg3 7.Kc2 Txg4 8.Pa3 Lg5
9.Kd3 Tg3+ 10.Kc4 Lf6 11.Pb1 b5+
12.Kb3 Kf7 13.Kc2 Tf3 14.Kb3 h5
15.Kc2 (223.937.857) 924

23/94 78:10 -14.87 2.Kf1 Lxb1 3.Pxb1 Td1+ 4.Ke2 Txb1
5.Kxe3 Td1 6.c4 a5 7.bxa5 bxa5 8.Ke2 Tc1
9.f4 a4 10.f5 a3 11.Kd2 Txc4 12.Kd3 Tc1
13.Kd2 Tc7 14.f6 gxf6 15.h4 (3.907.005.003) 832
User avatar
Mike S.
Posts: 1480
Joined: Thu Mar 09, 2006 5:33 am

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by Mike S. »

In the search of this particular position, Toga II 1.2.1a seemingly has a "fail high" (indicated by ++ in the output I get). I tried it again starting with the normal single-pv mode, and indeed there is something unusual. It doesn't seem to be able to finish the 16th ply which should result in the 10+ eval for 1...Bxb1.

Code: Select all

 14.01	 0:06 	+7.40 	1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Sxb1 Txf3 5.g4 Txh3 6.Ke2 Lf4 7.Kd1 Tg3 8.Kc2 Txg4 9.c4 Tg2+ 10.Kb3 Kf7 11.Sc3 h5 12.Se4 (5.421.582) 870.0
 15.01	 0:12 	+7.51 	1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Sxb1 Txf3 5.Kd1 Txg3 6.h4 Tg2 7.Sa3 Td2+ 8.Ke1 b5 9.h5 Ta2 10.Sb1 Th2 11.Kd1 Kf7 12.Sa3 Txh5 (10.633.978) 885.8
 15.03	 0:15 	+7.80++ 	1...Lxb1 2.Sxb1 Td1+ 3.Kf2 Txb1 4.gxf4 a5 5.bxa5 bxa5 6.Ke2 a4 7.c4 a3 8.c5 (13.715.831) 884.0
Bester Zug: La2xb1 Zeit: 7:25.000 min  K/s: 1.002.449  CPU 100.0%  Knoten: 446.090.000  
It switches to Bxb1 at depth 15 after 15 seconds, but after 7 minutes it still had not updated the output and was stuck at depth 16. But this is a very unusual case, or problem. It think it is not a question of single- or multi-pv mode. I am not aware that this is a frequent problem in Toga analysis, and I think it is a rare exception.

I switched to multi-pv mode. Seemingly due to remaining hashtable entries, it started with Bxb1 as the best pm immediatly and had no problem from depth 15 to 16:

Code: Select all

 10	 0:01 	+8.97 	1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 Txa3 5.cxb6 Tb3 6.Kf2 Txb4 7.Ke3 Lf5 8.h4 Txb6 (964.381) 920.0
 10	 0:01 	+7.20 	1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Sxb1 Txf3 5.Kd1 Txg3 6.h4 Tg2 7.Sa3 Td2+ 8.Ke1 Kf7 (999.056) 920.0
-----
(...)
-----
 15	 0:59 	+10.20 	1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 Txa3 5.cxb6 Tb3 6.Kf2 Txb4 7.Ke3 Lf5 8.Kd2 a5 9.Kc3 Lxh3 10.Kd2 a4 11.Kc3 Txb6 (47.361.834) 801.0
 15	 0:59 	+7.51 	1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Sxb1 Txf3 5.Kd1 Txg3 6.h4 Tg2 7.Sa3 Td2+ 8.Ke1 b5 9.h5 Ta2 10.Sb1 Th2 11.Kd1 Kf7 12.Sa3 Txh5 (47.565.521) 801.0
-----
 16	 2:24 	+10.77 	1...Lxb1 2.gxf4 Ld3 3.b5 axb5 4.h4 Ta2 5.Sxb5 Lxb5 6.h5 Ld3 7.Kh1 Kf7 8.Kg1 Ke6 9.Kh1 Kf5 10.Kg1 Kxf4 11.c4 Lxc4 (114.569.920) 793.8
LT:  15	 0:59 	+7.51 	1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Sxb1 Txf3 5.Kd1 Txg3 6.h4 Tg2 7.Sa3 Td2+ 8.Ke1 b5 9.h5 Ta2 10.Sb1 Th2 11.Kd1 Kf7 12.Sa3 Txh5 (47.565.521) 793.8
-----
 16	 2:24 	+10.77 	1...Lxb1 2.gxf4 Ld3 3.b5 axb5 4.h4 Ta2 5.Sxb5 Lxb5 6.h5 Ld3 7.Kh1 Kf7 8.Kg1 Ke6 9.Kh1 Kf5 10.Kg1 Kxf4 11.c4 Lxc4 (114.569.920) 802.1
 16	 2:39 	+7.58 	1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Sxb1 Txf3 5.h4 Txg3 6.Sa3 Lf4 7.Sc4 b5 8.Sb6 Txc3 9.Sd5 Lg3+ 10.Ke2 Tc2+ 11.Kd3 Tc4 12.Se7+ Kf8 13.Sd5 Lxh4 (128.349.501) 802.1
(outputs are taken from the Shredder 5 classic GUI in Windows)
Regards, Mike
User avatar
Eelco de Groot
Posts: 4660
Joined: Sun Mar 12, 2006 2:40 am
Full name:   Eelco de Groot

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by Eelco de Groot »

Output from Toga 1.2.1a seems comparable to what Richard is getting with single PV at 15 ply, here with 256 MB


### from Toga II 1.2.1a Default (0): info multipv 1 depth 15 seldepth 38 score cp 751 time 9265 nodes 10633978 pv f4e3 g1f1 d2f2 f1e1 a2b1 a3b1 f2f3 e1d1 f3g3 h3h4 g3g2 b1a3 g2d2 d1e1 b6b5 h4h5 d2a2 a3b1 a2h2 e1d1 g8f7 b1a3 h2h...
15/38 0:09 +7.51 1...Le3+ 2.Kf1 Tf2+ 3.Ke1 Lxb1 4.Pxb1 Txf3 5.Kd1 Txg3 6.h4 Tg2 7.Pa3 Td2+ 8.Ke1 b5 9.h5 Ta2 10.Pb1 Th2 11.Kd1 Kf7 12.Pa3 Txh5 (10.633.978) 1146
### from Toga II 1.2.1a Default (0): info currmove f4g3 currmovenumber 2


### from Toga II 1.2.1a Default (0): info multipv 1 depth 16 seldepth 86 score cp 1118 lowerbound time 1013812 nodes 1137471852 pv a2b1 g3f4 b1d3 c3c4 d2a2 c4c5 a2a3 c5b6 a3b3 g1f2 b3b4 f2e3 d3f5 b6b7 b4b7 e3d2 f5h3 f4f5 h3f5 f...
16/86 16:53 +11.18++ 1...Lxb1 2.gxf4 Ld3 3.c4 Ta2 4.c5 Txa3 5.cxb6 Tb3 6.Kf2 Txb4 7.Ke3 Lf5 8.b7 Txb7 9.Kd2 Lxh3 10.f5 Lxf5 11.f4 a5 12.Kc3 a4 13.Kc4 Le6+ 14.Kd3 (1.137.471.852) 1121
### from Toga II 1.2.1a Default (0): info depth 16 seldepth 86 time 1013812 nodes 1137471852 nps 1121975

So if Multi PV eval is increasing faster does it mean that Toga, Toga 1.2.1a in this case, is searching deeper at every plydepth in Multi PV mode? In that case does Multi PV not exactly what it should do for Richards purpose discriminating between different moves? Feature not a bug? Or is the search in single PV mode somehow not as efficient as it could be? Or is it a synergy effect in Multi PV mode. I know that that can happen especially when the two best moves transpose to the same variation.

In the analysis that I posted above I could not see this effect of faster rising multi PV eval with Toga II 1.3x4, at least not in this position with my settings?

-Probably a synergistic effect is there as well, seems logical, in case of a transposition at least it is similar to an SMP search with the simplest means of 'Search twice the same tree with shared hashtables between threads' method -

Eelco
Vempele

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by Vempele »

Eelco de Groot wrote:So if Multi PV eval is increasing faster does it mean that Toga, Toga 1.2.1a in this case, is searching deeper at every plydepth in Multi PV mode?
Yes. Toga searches PV nodes deeper because of the lack of history pruning (and the late move reduction thresholds are lower), and some extensions are only done in PV nodes. For example, 1.2.1a extends all captures in PV nodes as long as they aren't very far away from root.

This results in a more accurate search (with identical depth), but it takes longer. Increasing the number of Quiescence check plies is also useful in some tactical positions.
In that case does Multi PV not exactly what it should do for Richards purpose discriminating between different moves?
It wouldn't be too difficult to make Multi-PV do no extra work apart from finding out the second best move, but there'd be a big difference in accuracy between the best and second best move.
-Probably a synergistic effect is there as well, seems logical, in case of a transposition at least it is similar to an SMP search with the simplest means of 'Search twice the same tree with shared hashtables between threads' method -
Why would anyone want to search the same tree twice? Usually you just 'split' a position so that the threads don't search the same moves, even if you only use one split point (=all threads search the same position). That way, the trees are mostly separate.
rj909 wrote:Unfortunately TOGA II is sometimes way out on its scoring with MultiPV. Some positions when analyzed with standard options (i.e. no MultiPV analyze position after each move separately) the score for two alternatives may be very close (say <0.4). However in MultiPV mode the difference is >2.0 . I think this is a straight out bug as giving more time to MultiPV does not help (i.e. its not just that less of search space is explored due to processing the alternatives).
As I said, this is perfectly normal and not a bug. Rybka does the same. I'm not sure about Glaurung, but at least it does LMR in PV nodes as well, so the search could be a little more similar to single-PV.
rj909

Re: Bug with MultiPV in TOGA II - Alternatives with MultiPV?

Post by rj909 »

Thanks, I had assumed the multipv would be seeing less for each variation for a given amount of time than the single pv. You were right, Glaurung was closer to single-PV (as was Toga 1.3x4, I had only checked the MultiPV 2 on it previously so had assumed it had the same "problem" as Toga 1.2.1a).

Toga II 1.2.1a still looks like the odd one out (but I take your word for it that its not a bug).

Here are the results of very short runs (movetime 60 seconds) on my machine:

glaurung 1.2.1 Mac OS X Intel (Core Duo)

After 60 seconds.

info multipv 1 depth 19 score cp 1050 time 29221 nodes 27499806 nps 941097 pv a2b1 g3f4 b1d3 c3c4 d2a2 c4c5 b6c5 b4c5 a2a3 c5c6 a3c3 g1f2 d3f5 h3h4 a6a5 f2e1 a5a4 e1d1 c3c6


After 60 seconds multipv 2


info multipv 1 depth 19 score cp 1071 time 45240 nodes 43751346 nps 967094 pv a2b1 g3f4 b1d3 c3c4 d2a2 c4c5 b6c5 b4c5 a2a3 g1f2 d3b5 h3h4 b5c6 f2e1 a3f3 e1d2 f3f4 h4h5 a6a5
info multipv 2 depth 18 score cp 920 time 45241 nodes 43751346 nps 967072 pv f4e3 g1f1 d2f2 f1e1 a2b1 a3b1 f2f3 c3c4 e3d4 g3g4 f3h3 b1d2 d4c3 b4b5 a6b5 e1d1 c3d2 d1d2 b5c4 d2c2


Toga II 1.2.1a Mac OS X Intel (Core Duo)

After 60 seconds

info multipv 1 depth 15 seldepth 47 score cp 780 lowerbound time 17658 nodes 14512450 pv a2b1 a3b1 d2d1 g1f2 d1b1 g3f4 a6a5 b4a5 b6a5 f2e2 a5a4 c3c4 a4a3 c4c5

After 60 seconds MultiPv 2

info multipv 1 depth 14 seldepth 58 score cp 966 time 19103 nodes 14840054 pv a2b1 g3f4 b1d3 b4b5 d2a2 b5a6 a2a3 a6a7 a3a7 g1f2 a7c7 f2e3 d3f5 e3d4 f5h3 d4d3 h3g2 c3c4 g2f1 d3e3 f1c4
info multipv 2 depth 14 seldepth 58 score cp 741 time 26044 nodes 20490943 pv f4e3 g1f1 d2f2 f1e1 a2b1 a3b1 f2f3 g3g4 f3h3 e1d1 h3g3 b1a3 g3g4 d1c2 h7h5 c2d3 g4g3 a3c4 e3g5 d3c2 g5f6 c4b6 g3c3 c2d2

Toga II 1.3x4 Mac OS X Intel (Core Duo)

After 60 seconds
info multipv 1 depth 17 seldepth 50 score cp 979 time 23905 nodes 15517098 pv a2b1 g3f4 b1d3 b4b5 d2a2 b5a6 a2a3 a6a7 a3a7 g1f2 a7c7 f2e3 d3f5 e3d4 f5h3 d4d3 h3g2 f4f5 g2f3 d3d4 c7d7 d4e5

After 60 seconds MultiPv 2
info multipv 1 depth 17 seldepth 44 score cp 985 time 33729 nodes 22729637 pv a2b1 g3f4 b1d3 b4b5 d2a2 b5a6 a2a3 a6a7 a3a7 g1f2 d3f5 h3h4 a7c7 f2e3 c7c3 e3d4 c3f3 d4e5 f5g6 h4h5 g6h5
info multipv 2 depth 17 seldepth 53 score cp 761 time 47145 nodes 31815386 pv f4e3 g1f1 d2f2 f1e1 a2b1 a3b1 f2f3 h3h4 f3g3 e1d1 g3h3 d1c2 a6a5 b1a3 h3h4 c2d3 h4h3 a3c4 e3f2 d3d2 a5a4 d2c2