Obsidian can be fast in test positions

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

Moderator: Ras

peter
Posts: 3358
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Obsidian can be fast in test positions

Post by peter »

Sorry, not sure anymore, if the output above really was single threaded, guess not, so I repeated the trial, here's another one, now done with certainly 1 thread only, of course after restart of GUI again:

Code: Select all

rkr5/p5p1/Pp4Kp/8/3B4/1B6/8/8 w - - 0 1

Analysis by Obsidian 15.0 JA:

1.Le5+ Tc7 2.Ld5 Kc8 3.Lxc7 Kxc7 4.Lxa8 b5 5.Kxg7 h5 6.Kf6 h4 7.Kg5 h3 
  Schwarz hat minimalen Vorteil.  
  = (-0.30)  Tiefe: 10   00:00:00
...
1.Le5+ Tc7 2.Ld5 b5 3.Lxa8 b4 4.Lb7 b3 5.Ld5 Kc8 6.Le6+ Kd8 7.Lxc7+ Kxc7 8.Lxb3 h5 9.Kxh5 
  Die Stellung ist ausgeglichen.  
  = (0.00)  Tiefe: 64   00:00:08  16142kN, tb=559204
1.Le5+ Tc7 2.Le6 b5 3.Ld7 b4 4.Kf7 b3 5.Ke7 b2 6.Kd8 b1D 7.Lxc7# 
  Weiß setzt Matt.  
  +- (#7)  Tiefe: 65   00:00:11  24562kN, tb=584913
1.Le5+ Tc7 2.Le6 b5 3.Ld7 b4 4.Kf7 b3 5.Ke7 b2 6.Kd8 b1D 7.Lxc7# 
  Weiß setzt Matt.  
  +- (#7)  Tiefe: 66   00:00:12  34641kN, tb=589898
...
1.Le5+ Tc7 2.Le6 b5 3.Ld7 b4 4.Kf7 b3 5.Ke7 b2 6.Kd8 b1D 7.Lxc7# 
  Weiß setzt Matt.  
  +- (#7)  Tiefe: 123   00:00:39  250MN, tb=707145
Peter.
Uri Blass
Posts: 10768
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Obsidian can be fast in test positions

Post by Uri Blass »

for comparison chest a mate solver with 1 thread needed 0.2 seconds to see the mate so it seems obsidian is not so fast.

FEN: rkr5/p5p1/Pp4Kp/8/3B4/1B6/8/8 w - - 0 1

ChestUCI:
FEN: rkr5/p5p1/Pp4Kp/8/3B4/1B6/8/8 w - - (4+7)
Stellungs-Analyse: C0/R0/K2/P4/X15 W:3/25
Problem in Datenbank gefunden: #7; 00:00;
Suche nach Matt in 126 ... (Hash=2044MB)
7 00:00 0 0 +M7 Bd4-e5+
Suche abgeschlossen ... (Zeit=0.20s)
Matt in 7 gefunden ! (1 Lösung in 00:00)
7/7 00:00 171k 845k +M7 Bd4-e5+ Rc8-c7 Bb3-e6 h6-h5 Be6-d7 h5-h4 Kg6-f7 h4-h3 Kf7-e7 h3-h2 Ke7-d8 h2-h1Q Be5xc7+
peter
Posts: 3358
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Obsidian can be fast in test positions

Post by peter »

Uri Blass wrote: Mon Mar 10, 2025 11:59 am for comparison chest a mate solver with 1 thread needed 0.2 seconds to see the mate so it seems obsidian is not so fast.
Well, Uri, but that's not a fair comparison, Chest is a specialised mate solving program, not a game playing one, which Obsidian yet and successful is, regards
Peter.
Uri Blass
Posts: 10768
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Obsidian can be fast in test positions

Post by Uri Blass »

peter wrote: Mon Mar 10, 2025 12:04 pm
Uri Blass wrote: Mon Mar 10, 2025 11:59 am for comparison chest a mate solver with 1 thread needed 0.2 seconds to see the mate so it seems obsidian is not so fast.
Well, Uri, but that's not a fair comparison, Chest is a specialised mate solving program, not a game playing one, which Obsidian yet and successful is, regards

Not always chest is the fastest in solving mates(there are cases when normal engines ate faster because they do not need to prove that there is no shorter mate).

In any case there are other playing engines that are faster than Obsidian.

FEN: rkr5/p5p1/Pp4Kp/8/3B4/1B6/8/8 w - - 0 1

Berserk13:
time -1 start 1180417921 alloc 0 max 2147483647 depth 200 timeset 0 searchmoves 0
1/1 00:00 18 18k -0.41 Bd4-e5+
2/2 00:00 42 42k -0.41 Bd4-e5+ Rc8-c7
3/3 00:00 74 74k -0.79 Bd4-e5+ Rc8-c7 Be5-d6
4/4 00:00 123 123k -1.33 Bd4-e5+ Rc8-c7 Be5-d6 Kb8-c8
5/5 00:00 192 192k -0.92 Bd4-e5+ Rc8-c7 Be5-g3 Kb8-c8 Bb3-e6+
6/6 00:00 279 279k -0.92 Bd4-e5+ Rc8-c7 Be5-g3 Kb8-c8 Bb3-e6+ Rc7-d7
7/9 00:00 486 486k -0.12 Bd4-e5+
8/7 00:00 804 804k -0.09 Bd4-e5+ Rc8-c7 Bb3-d5 h6-h5 Bd5xa8 Kb8-c8 Be5xc7
9/11 00:00 2k 2,336k -0.18 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7
10/12 00:00 2k 2,480k -0.17 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 h6-h5 Ba8-g2 b5-b4
11/15 00:00 4k 4,264k -0.07 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Kg7xh6
12/14 00:00 5k 4,554k -0.11 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Kg7xh6 Kb6xa6 Kh6-g6 Ka6-b6
13/19 00:00 8k 8,300k -0.09 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Kg7xh6 Kb6xa6 Kh6-g6 Ka6-b6 Kg6-f5 b5-b4
14/20 00:00 12k 12,233k -0.09 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 h6-h5 Kg7-g6 Kc7-b6 Kg6xh5 Kb6xa6 Kh5-g4 Ka6-b6 Kg4-f5 a7-a6
15/21 00:00 18k 17,861k -0.06 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 Kc7-b6 Kg4xh3 Kb6xa6
16/26 00:00 64k 2,010k -0.28 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-b5 Kf4-e3 a7-a5
17/25 00:00 73k 2,296k -0.31 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-b5 Be4-d3+ Kb5-b4 Kf4-e3 a7-a5 Ke3-d4
18/29 00:00 87k 1,841k -0.16 Bd4-e5+ Rc8-c7 Bb3-d5 b6-b5 Bd5xa8 Kb8-c8 Be5xc7 Kc8xc7 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-b5 Kf4-e3 Kb5-b4
19/25 00:00 100k 2,138k -0.15 Bd4-e5+ Rc8-c7 Bb3-d5 b6-b5 Bd5xa8 Kb8-c8 Be5xc7 Kc8xc7 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-b5 Kf4-e3 Kb5-b4 Ke3-d2 Kb4-a3 Kd2-c3
20/28 00:00 119k 1,886k -0.11 Bd4-e5+ Rc8-c7 Bb3-d5 b6-b5 Bd5xa8 Kb8-c8 Be5xc7 Kc8xc7 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-b5 Kf4-e3 Kb5-b4 Ke3-d2 Kb4-a3 Be4-c2 Ka3-a2 Kd2-c3 a7-a5
21/31 00:00 181k 1,920k -0.12 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d2 Kb4-a3 Kd2-c3 Ka3-a2
22/29 00:00 211k 1,916k -0.12 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d2 Kb4-a3 Kd2-c3 Ka3-a2 Kc3-b4 b2-b1Q+ Be4xb1+
23/32 00:00 231k 2,104k -0.12 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d2 Kb4-a3 Kd2-c3 Ka3-a2 Kc3-b4 b2-b1Q+ Be4xb1+ Ka2xb1 Kb4-c4 Kb1-a1
24/35 00:00 306k 1,950k -0.14 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 b5-b4 Kg7xh6 b4-b3 Kh6-g5 b3-b2 Bb7-e4 Kb6xa6 Kg5-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d2 Kb4-a3 Kd2-c3 Ka3-a2 Kc3-b4 b2-b1Q+ Be4xb1+ Ka2xb1 Kb4-a3 Kb1-a1 Ka3-b3 a7-a5 Kb3-c3
25/39 00:00 335k 1,947k -0.08 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 h3-h2 Kg4-g3 h2-h1Q Bb7xh1 Kb6xa6 Kg3-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d4 a7-a5 Bh1-e4
26/32 00:00 385k 2,045k -0.08 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 h3-h2 Kg4-g3 h2-h1Q Bb7xh1 Kb6xa6 Kg3-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d4 a7-a5 Bh1-a8 a5-a4 Ba8-d5 Kb4-a3
27/35 00:00 449k 2,052k -0.06 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 h3-h2 Kg4-g3 h2-h1Q Bb7xh1 Kb6xa6 Kg3-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d4 a7-a5 Bh1-a8 Kb4-b3 Kd4-c5 a5-a4 Ba8-d5+ Kb3-b2
28/40 00:00 579k 2,055k -0.03 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 h3-h2 Kg4-g3 h2-h1Q Bb7xh1 Kb6xa6 Kg3-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d4 a7-a5 Bh1-d5 a5-a4 Bd5-a8 Kb4-b3 Ba8-d5+ Kb3-b2 Kd4-c5 b5-b4 Kc5xb4
29/41 00:00 716k 2,082k -0.04 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 h3-h2 Kg4-g3 h2-h1Q Bb7xh1 Kb6xa6 Kg3-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d4 a7-a5 Bh1-c6 Kb4-b3 Bc6xb5 Kb3-b4 Bb5-e8 a5-a4 Be8-g6 Kb4-b3
30/43 00:00 842k 2,153k -0.04 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 h3-h2 Kg4-g3 h2-h1Q Bb7xh1 Kb6xa6 Kg3-f4 Ka6-a5 Kf4-e3 Ka5-b4 Ke3-d4 Kb4-b3 Bh1-d5+ Kb3-a3 Kd4-c3 a7-a5 Bd5-e6 b5-b4+ Kc3-c2 a5-a4 Kc2-b1 b4-b3 Kb1-a1
31/45 00:00 1,036k 2,136k -0.06 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 h3-h2 Kg4-g3 h2-h1Q Bb7xh1 Kb6xa6 Kg3-f4 Ka6-b6 Kf4-e3 Kb6-c5 Bh1-f3 a7-a5 Ke3-d3 a5-a4 Bf3-h5 Kc5-b4 Kd3-c2 Kb4-a3 Bh5-e2 Ka3-a2 Kc2-c3
32/42 00:00 1,161k 2,182k -0.04 Bd4-e5+ Rc8-c7 Bb3-d5 Kb8-c8 Be5xc7 Kc8xc7 Bd5xa8 b6-b5 Kg6xg7 Kc7-b6 Ba8-b7 h6-h5 Kg7-g6 h5-h4 Kg6-g5 h4-h3 Kg5-g4 h3-h2 Kg4-g3 h2-h1Q Bb7xh1 Kb6xa6 Kg3-f4 Ka6-b6 Kf4-e3 Kb6-c5 Bh1-f3 a7-a5 Ke3-d3 Kc5-d6 Kd3-c3 Kd6-e6 Kc3-b3 Ke6-f7 Bf3-d5+ Kf7-g7 Bd5-e4
33/41 00:00 1,442k 2,249k +M9 Bd4-e5+ Rc8-c7 Bb3-e6 b6-b5 Be6-d7 b5-b4 Kg6-f7 h6-h5 Kf7-e7 g7-g6 Ke7-e8 g6-g5 Ke8-e7 b4-b3 Ke7-d8 b3-b2 Be5xc7+
34/18 00:00 1,468k 2,235k +M7 Bd4-e5+ Rc8-c7 Bb3-e6 b6-b5 Be6-d7 b5-b4 Kg6-f7 h6-h5 Kf7-e7 g7-g6 Ke7-d8 b4-b3 Be5xc7+
peter
Posts: 3358
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Obsidian can be fast in test positions

Post by peter »

Uri Blass wrote: Mon Mar 10, 2025 12:54 pm
peter wrote: Mon Mar 10, 2025 12:04 pm
Uri Blass wrote: Mon Mar 10, 2025 11:59 am for comparison chest a mate solver with 1 thread needed 0.2 seconds to see the mate so it seems obsidian is not so fast.
Well, Uri, but that's not a fair comparison, Chest is a specialised mate solving program, not a game playing one, which Obsidian yet and successful is, regards

Not always chest is the fastest in solving mates(there are cases when normal engines ate faster because they do not need to prove that there is no shorter mate).

In any case there are other playing engines that are faster than Obsidian.
Well done, Berserk!
Yet e.g. Stockfish dev. (20250202 is the latest one I downloaded so far) takes some longer single threaded. CorChess (less nullmove pruning) does bettter.
And no, it's not a really hard one test position, especially if you count finding best move as time to solution already, e.g. Lc0 0.31.2 with net 6147500PT and 3070ti GPU has a DTM- output at once too, just takes a little longer to come from #8 to correct #7.
I just also tried this one position, because Marek quoted it in his thread about new Sting.
Huntsman of course is at once at #7 single threaded with MateSearch checked too, regards
Peter.
lech
Posts: 1167
Joined: Sun Feb 14, 2010 10:02 pm

Re: Obsidian can be fast in test positions

Post by lech »

lech wrote: Sun Mar 09, 2025 3:57 pm If a search results in checkmate, the best advice is to stop and restart the search immediately. This should work fine for all chess engines if someone wants to get the shortest path to checkmate quickly.
The issue of finding the shortest checkmate (when a position is solved) seems to be a very peripheral problem and I don't know why engine developers should take it seriously. Of course, Hash is not cleaning before restart of solving. It is not a trick (help). It saves time and energy only.

BTW. I tried Sting-sf-bh-10 (1 thread, Hash 1024) with the above Popov's #9 and it works very fine.
Maybe, I can't be friendly, but let me be useful.
lech
Posts: 1167
Joined: Sun Feb 14, 2010 10:02 pm

Re: Obsidian can be fast in test positions

Post by lech »

peter wrote: Mon Mar 10, 2025 1:02 pm --- I just also tried this one position, because Marek quoted it in his thread about new Sting. ---
Some problems can be considered studies. The Mendheim's #7, without the dual 4. Kg7, would be a good example. Some time ago many strong engines (ELO) were unable to solve it.
Maybe, I can't be friendly, but let me be useful.