Test Position---Not For Over Priced Commercials

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

Moderator: Ras

kgburcham
Posts: 2016
Joined: Sun Feb 17, 2008 4:19 pm

Test Position---Not For Over Priced Commercials

Post by kgburcham »

17.Rxf6

[d] r1b1k2r/1p1n1pp1/p2Qpn2/8/P2B3p/1qN3P1/1P2P1BP/R3KR2 w kq -
Top
User avatar
Eelco de Groot
Posts: 4671
Joined: Sun Mar 12, 2006 2:40 am
Full name:   Eelco de Groot

Re: Test Position---Not For Over Priced Commercials

Post by Eelco de Groot »

That looks like a good testmove! Pretty difficult for the fruitoga-varieties that I tried. Probably a real tactician does much better here, Deep Fritz 11 or Winfinder? With Toga Checkov in the previous to last builds I could not find the move, only after switching to Multi-PV after a normal search so the hashtables had a lot of results from the normal search, Rxf6 was found fast. Probably a combination of NullMove pruning and 'Brute Pruning' in non-PV nodes was too much, in Multi-PV mode more nodes are treated as PV nodes.

But there are a lot of ways to bore out the cilinders of the new single reply extensions so I tried a new variation of that, in Build 135. I don't think it is correct, it even refuses to take the Rook on f6 but it does find 1.Rxf6!

Build 134 plays 1. Rd1 with a good score but Build 135 leaves that variation quickly, whether for the right reasons or not, incorrect or not I'm not really sure.

r1b1k2r/1p1n1pp1/p2Qpn2/8/P2B3p/1qN3P1/1P2P1BP/R3KR2 w kq -

Engine: Toga 1.4 Checkov Beta 4/Toga Mara Beta 6 Build 134 (Athlon 2009 MHz, 256 MB)
by Fabien Letouzey, Thomas Gaksch, Taban Juma, Eelco de Groot

8/46 0:04 +1.47 1.Rd1 e5 2.Nd5 Nxd5 3.Bxd5 Rh6
4.Qxh6 Qxd5 5.Qg5 hxg3 6.hxg3 (2.576.194) 552

9/46 0:10 +0.68 1.Rd1 Rh5 2.gxh4 Rxh4 3.Qc7 Rh6
4.Rd2 Ng4 5.h3 Ngf6 6.Bxf6 Rxf6 (5.673.998) 562

9/46 0:14 +1.14 1.Rf3 Qc2 2.Rd1 Rh6 3.Bc5 Ng8 4.Be3 Rh5
5.g4 Qxb2 (8.121.626) 557

10/46 0:17 +1.02 1.Rf3 Qc2 2.Rd1 h3 3.Bc5 Ng8 4.Bf1 Qxb2
5.Rd2 Qb3 6.a5 (9.522.935) 555

10/52 0:22 +1.44 1.Rd1 h3 2.Be4 Rh5 3.Rxf6 gxf6
4.Bxf6 Qxd1+ 5.Kxd1 Nxf6 6.Bc6+ Nd7
7.g4 Rg5 (12.510.752) 559

11/58 0:47 +1.38 1.Rd1 Rh6 2.Bc5 Ng8 3.Qc7 hxg3 4.Ne4 e5
5.hxg3 Ne7 6.Nd6+ Rxd6 7.Bxd6 Qxg3+
8.Rf2 Nf5 (27.344.594) 575

12/58 1:29 +2.02 1.Rd1 hxg3 2.Rd3 Qc4 3.b3 Qxb3 4.Nd5 Qb1+
5.Rd1 Qxd1+ 6.Kxd1 exd5 7.Qxg3 Rh6
8.Qf4 Kd8 9.Qd6 Ne4 10.Qxd5 (52.102.554) 582

13/58 3:54 +2.12 1.Rd1 h3 2.Bf3 Rh6 3.Rd3 Qc2 4.Bc5 Ng8
5.Bxb7 Qc1+ 6.Kf2 Rf6+ 7.Rf3 Rxf3+
8.Bxf3 Qxb2 9.Bxa8 Qxc3 (137.784.607) 588

14/58 6:52 +2.46 1.Rd1 h3 2.Bh1 Rh5 3.Rd3 Ng4 4.Nb5 Rxb5
5.Rxb3 Rxb3 6.Rf3 Rxf3 7.exf3 e5
8.Bg1 Ngf6 9.f4 exf4 10.Qxf4 (243.883.454) 590

15/69 18:57 +2.51 1.Rd1 Rh5 2.gxh4 e5 3.Nd5 Nxd5
4.Bxd5 Rh6 5.Qxh6 Qxd1+ 6.Kxd1 gxh6
7.Be3 f6 8.Bxh6 Nb6 9.Bb3 Bg4
10.Rxf6 O-O-O+ 11.Bd2 Kc7 12.Rf7+ Kc6 (665.534.802) 585

16/69 31:46 +2.56 1.Rd1 Rh5 2.gxh4 Rb8 3.Qc7 Rxh4
4.Bxf6 gxf6 5.Rxd7 Bxd7 6.Qxb8+ Ke7
7.Qc7 Rc4 8.Qxb7 Qxb7 9.Bxb7 a5
10.Ba6 Rb4 11.Bb5 Rxb2 12.Bxd7 Kxd7
13.Rxf6 (1.111.039.729) 582

1. Rxf6 is not found within half an hour

New adaptation of Singular Extensions in Build 135 creating a Coffee House version of Toga :) :

[d]r1b1k2r/1p1n1pp1/p2Qpn2/8/P2B3p/1qN3P1/1P2P1BP/R3KR2 w kq -

Engine: Toga 1.4 Checkov Beta 4/Toga Mara Beta 6 Build 135 (Athlon 2009 MHz, 256 MB)
by Fabien Letouzey, Thomas Gaksch, Taban Juma, Eelco de Groot

8/46 0:04 +1.47 1.Rd1 e5 2.Nd5 Nxd5 3.Bxd5 Rh6
4.Qxh6 Qxd5 5.Qg5 hxg3 6.hxg3 (2.576.173) 557

9/46 0:10 +1.48 1.Rd1 Rh5 2.Rd3 Qxb2 3.Nb5 Qb1+
4.Rd1 Rxb5 5.Rxb1 Rxb1+ 6.Kf2 Ng4+
7.Kg1 Rxf1+ 8.Bxf1 hxg3 9.hxg3 (5.873.155) 562

10/46 0:16 +1.42 1.Rd1 h3 2.Be4 Rh6 3.Bc5 Ng8 4.Be3 Rh5
5.Rd2 f5 (9.448.298) 558

11/46 0:30 +1.24 1.Rd1 h3 2.Be4 Rh5 3.Rxf6 gxf6
4.Bxf6 Qxd1+ 5.Kxd1 Nxf6 6.Bf3 Rh8
7.Qf4 Nd7 (17.239.748) 560

12/50 1:14 +2.15 1.Rd1 hxg3 2.Rd3 Qc4 3.b3 Qxb3 4.Nd5 Qb1+
5.Rd1 Qxd1+ 6.Kxd1 Nxd5 7.Bxd5 exd5
8.Qxg3 Rh7 9.Kc2 g6 10.Qe3+ Kf8 (42.639.747) 576

13/60 3:15 +1.67 1.Rd1 Rh5 2.Rd3 Qxb2 3.gxh4 Rh6
4.Nb5 Qb1+ 5.Rd1 Qxd1+ 6.Kxd1 axb5
7.axb5 Rxh4 8.Bc3 Rh5 9.b6 Ra2 (114.147.942) 583

13/60 5:15 +2.53 1.Rxf6 gxf6 2.Nd5 exd5 3.Ra3 Qc4
4.Re3+ Ne5 5.Bxe5 fxe5 6.Qxe5+ Kd7
7.Rc3 Rd8 8.Rxc4 dxc4 9.gxh4 a5
10.Kf2 f5 11.Qxf5+ Ke7 12.Qe4+ Kd7 (189.529.174) 601

14/60 7:39 +4.59 1.Rxf6 gxf6 2.Ne4 Rh6 3.Nd2 Qc2
4.Be4 hxg3 5.h3 Rxh3 6.Bxc2 e5 7.Rd1 exd4
8.Qxd4 f5 9.Nb3 a5 10.Bxf5 Rh1+
11.Kd2 (280.406.093) 610

15/60 16:08 +4.82 1.Rxf6 a5 2.Ne4 Rh5 3.Rf3 Qd5 4.g4 Ra6
5.Qxd5 Rxd5 6.Bxg7 Ke7 7.Rd1 Rc6
8.Rxd5 exd5 9.Nc3 Nb6 (593.921.029) 613

16/64 28:01 +4.91 1.Rxf6 a5 2.Rf1 hxg3 3.Qxg3 Rg8
4.Bxg7 Qxb2 5.Rb1 Qa3 6.Bf3 e5 7.Nd5 Qc5
8.Qg5 b6 9.Rb5 Qd6 10.Nf6+ Qxf6
11.Qxf6 Nxf6 12.Rxe5+ Be6 13.Bxa8 (1.039.515.358) 618

17/76 42:46 +5.00 :) 1.Rxf6 a5 2.Nb5 Qb4+ 3.Qxb4 axb4
4.Nc7+ Kd8 5.Nxa8 gxf6 6.Nb6 hxg3
7.hxg3 Rh2 8.Kf1 Nxb6 9.Bxb6+ Ke8
10.Rc1 Bd7 11.a5 Bb5 12.Bxb7 Kd7 (1.585.399.654) 617


Eelco

Code: Select all


  // [EdG: single reply extensions hack]
    
   if (best_move != MoveNone && nextbest_value != ValueNone && node_type == NodePV && best_value < beta
	   && depth_best_move < depth && ((best_value - nextbest_value) >= (300 - 10 * depth)) && depth <= 30){// [EdG:[]'only move' hack]
	  move_do(board,best_move,undo);
	  value = -full_search(board,-beta,-alpha,depth_best_move + 1,height+1,new_pv,NodePV,true,ThreadId);
	  //value = -full_search(board,-beta,-alpha,depth_best_move + 2,height+1,new_pv,NodePV,cap_extended,ThreadId);
	  //value = -full_search(board,-beta,-alpha,depth_best_move + 3,height+1,new_pv,NodePV,cap_extended,ThreadId);
	  move_undo(board,best_move,undo);
	  
	  best_value = value;
	  pv_cat(pv,new_pv,best_move);

	  ASSERT(value_is_ok(best_value));

    if (value > alpha) {
       alpha = value;
    }
	  if (value < nextbest_value) { // This part is new, extension of nextbest_move when best_move fails low, untested so far.

		 move_do(board,nextbest_move,undo);
       nextbest_value = -full_search(board,-beta,-alpha,depth_best_move + 1,height+1,new_pv,NodePV,true,ThreadId);
       move_undo(board,nextbest_move,undo);
		 if (value < nextbest_value) {
         best_value = nextbest_value;
		   pv_cat(pv,new_pv,nextbest_move);
		    		 
		   ASSERT(value_is_ok(best_value));

         if (best_value > alpha) {
            alpha = best_value;
         }
		 } 
	  }

	  /*
	  if (old_alpha > value && nextbest_value > value) {
		 alpha = old_alpha;
		 best_move = nextbest_move;
		 best_value = nextbest_value;
		 pv_cat(pv,new_pv,nextbest_move);
	  }
	  */
   } // end of single reply extensions

cut:
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: 4671
Joined: Sun Mar 12, 2006 2:40 am
Full name:   Eelco de Groot

Re: Test Position---Not For Over Priced Commercials

Post by Eelco de Groot »



19/79 82:33 +5.37 :) 1.Rxf6 a5 2.Nb5 Qb4+ 3.Qxb4 axb4
4.Nc7+ Ke7 5.Rxf7+ Kxf7 6.Nxa8 hxg3
7.hxg3 e5 8.Bg1 Ke6 9.Rd1 Kf5 10.Rd6 Rh5
11.a5 b5 12.Nc7 (3.051.788.698) 616


I have detected a bug in my code though, best_move is not updated after a fail low that goes below the extended nextbest_move, so hashmoves and moveordering are probably not correct anymore in such cases..
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
swami
Posts: 6662
Joined: Thu Mar 09, 2006 4:21 am

Re: Test Position---Not For Over Priced Commercials

Post by swami »

r1b1k2r/1p1n1pp1/p2Qpn2/8/P2B3p/1qN3P1/1P2P1BP/R3KR2 w kq - 0 1

Analysis by Fruit 05/11/03:

Code: Select all

1. +-  (3.13): 1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7 5.Bc5 0-0-0 6.Bb6 Qc4 7.Qe7 Kb8 8.Qxd8+ Qc8 9.Nxf6 Qxd8 10.Bxd8 Bf5 
2. +-  (3.12): 1.Rd1 Rh5 2.g4 h3 3.Bxf6 gxf6 4.Bxb7 Qxb7 5.gxh5 Qb6 6.Qxb6 Nxb6 7.Rxf6 Bd7 8.Rd4 Ke7 9.Rf3 Rg8 10.Rxh3 
3. +-  (2.82): 1.g4 h3 2.Bh1 e5 3.Rxf6 Nxf6 4.Nd5 Nxd5 5.Bxd5 Qxd5 6.Qxd5 exd4 7.Qe5+ Kf8 8.Qxd4 Be6 9.Qb4+ Kg8 10.Qxb7 Rd8 
4. +-  (1.74): 1.Rf3 h3 2.Nb5 axb5 3.Rxb3 hxg2 4.Bg1 bxa4 5.Qd4 Nb8 6.Rba3 Nc6 7.Qd3 Ng4 8.Qf3 Nxh2 9.Qxg2 
5. +/-  (1.01): 1.Rf4 hxg3 2.Bxf6 Nxf6 3.Rb4 Qc2 4.Rd1 Bd7 5.Rxb7 Rd8 6.Rxd7 Nxd7 7.Bc6 Qxd1+ 8.Nxd1 gxh2 9.Nf2 g5 10.Kf1 g4 
6. +/-  (1.00): 1.Ne4 Nxe4 2.Bxe4 a5 3.Bxg7 Rg8 4.Bc3 hxg3 5.hxg3 Qc4 6.Bd3 Qc6 7.Qf4 f5 8.Qh4 Nc5 9.Qh5+ Kd8 
7. +/=  (0.68): 1.Rc1 hxg3 2.hxg3 Rh6 3.Rd1 e5 4.Bc5 Qe6 5.Qxe6+ fxe6 6.Be3 Rh2 7.Be4 Rh3 8.Bg6+ Ke7 9.Ne4 Rh2 10.Nd6 
8. +/-  (0.74): 1.Rf2 Rh5 2.e3 Rh6 3.Rd1 e5 4.Bc5 Qe6 5.Qc7 e4 6.Rfd2 hxg3 7.hxg3 Rh2 8.Bb4 
9. +/-  (0.74): 1.Bh1 hxg3 2.Rd1 Rh5 3.Rxf6 gxf6 4.Bxf6 Qxd1+ 5.Kxd1 Nxf6 6.Qxg3 Rf5 7.Ne4 Rd5+ 8.Ke1 Nxe4 9.Bxe4 Rc5 10.h4 
10. +/=  (0.70): 1.Rb1 hxg3 2.hxg3 e5 3.Bd5 Qc2 4.Bxe5 Rh2 5.Bxf6 Nxf6 6.Rd1 Nxd5 7.Qxd5 Be6 8.Qxb7 Rd8 9.Qc6+ Rd7 10.Rxd7 Bxd7 
11. +/=  (0.65): 1.gxh4 Rxh4 2.Rd1 Rxd4 3.Rxd4 Qxb2 4.Rf3 Qb6 5.Qxb6 Nxb6 6.Rg3 Kf8 7.Rg5 Rb8 8.a5 Nbd7 9.Kf2 
12. +/=  (0.44): 1.e3 hxg3 2.hxg3 Rh6 3.Rd1 e5 4.Bc5 Ng8 5.Qd2 Rh2 6.Ne4 Rxg2 7.Qxg2 Nxc5 8.Nd6+ Kd8 9.Nxf7+ Kc7 
13. +/=  (0.43): 1.Ra3 Qc2 2.Nd1 hxg3 3.Rxg3 Qh7 4.Qc7 Qh4 5.Rf4 Qh7 6.Nc3 Qc2 7.Rxg7 Nh5 8.Rh4 Qc1+ 9.Kf2 Qxb2 
14. +/=  (0.43): 1.a5 hxg3 2.hxg3 e5 3.Nd5 Nxd5 4.Bxd5 Qxg3+ 5.Kd1 f6 6.Bf2 Qh3 7.Rh1 Qxh1+ 8.Bxh1 Rxh1+ 9.Be1 Nf8 10.Qc7 Ne6 
15. +/=  (0.36): 1.h3 e5 2.Rxf6 exd4 3.Bd5 Qb6 4.Bxf7+ Kd8 5.Qxb6+ Nxb6 6.Rxb6 dxc3 7.g4 Kc7 8.Rb3 cxb2 9.Rxb2 Bd7 10.Rc1+ Bc6 11.e4 Rad8 
16. +/=  (0.32): 1.Rh1 hxg3 2.Rd1 Rxh2 3.Rxh2 gxh2 4.Rd3 Qc2 5.Rh3 Nh7 6.Rxh2 g5 7.Be4 Qc1+ 8.Nd1 Nhf8 9.Rh8 
17. =  (0.21): 1.Bf3 hxg3 2.hxg3 e5 3.Nd5 Nxd5 4.Bxd5 Qxg3+ 5.Kd1 Rf8 6.Kd2 Qg5+ 7.Be3 Qh4 8.Qc7 Qb4+ 9.Kc2 f5 10.Kb1 
18. =  (0.15): 1.Nd1 a5 2.Ne3 hxg3 3.hxg3 Rh2 4.Rf2 Rh8 5.Bc3 Rh6 6.Kf1 Qb6 7.Nc4 Qxd6 8.Nxd6+ Ke7 
19. =  (0.00): 1.e4 e5 2.Bxe5 Rh6 3.Bd4 Qxb2 4.Ra2 Qb3 5.Ra3 Qc2 6.Ra2 Qb3
Top
User avatar
Ovyron
Posts: 4562
Joined: Tue Jul 03, 2007 4:30 am

Re: Test Position---Not For Over Priced Commercials

Post by Ovyron »

Eelco de Groot wrote:Probably a real tactician does much better here, Deep Fritz 11 or Winfinder?
Winfinder doesn't find it in reasonable time. The problem may be that g4 also wins and it's not designed to find the "best win."
Top
kgburcham
Posts: 2016
Joined: Sun Feb 17, 2008 4:19 pm

Old Deep Fritz 6

Post by kgburcham »

[d] r1b1k2r/1p1n1pp1/p2Qpn2/8/P2B3p/1qN3P1/1P2P1BP/R3KR2 w kq - 0 1

Deep Fritz 6:

1.Rxf6!
+- (1.59) Depth: 13/37 00:00:59 368mN
1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7 5.Bc5 0-0-0 6.Bb6 Qc4
+- (2.00) Depth: 13/37 00:01:04 398mN
1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7 5.Bc5 0-0-0 6.Bb6 Qc4
+- (2.00) Depth: 14/36 00:01:11 448mN
1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7 5.Bc5 0-0-0 6.Bb6 Qc4
+- (2.09) Depth: 15/39 00:02:38 1039mN
1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7 5.Bc5 0-0-0 6.Bb6 Qc4
+- (2.22) Depth: 16/41 00:05:58 2369mN
1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7 5.Bc5 0-0-0 6.Bb6 Qc4
+- (2.31) Depth: 17/45 00:12:59 5300mN
Top
User avatar
AdminX
Posts: 6363
Joined: Mon Mar 13, 2006 2:34 pm
Location: Acworth, GA

Re: Test Position---Not For Over Priced Commercials

Post by AdminX »

Crafty likes Rd1 even better than Rxf6.

Analysis by Crafty 22.5:

1.Ne4 Nxe4 2.Bxe4 hxg3 3.hxg3 Rh2 4.Bxg7 f5 5.Rh1
+/= (0.34) Depth: 7 00:00:00 323kN
= (0.19) Depth: 9 00:00:00 1082kN

1.gxh4 Rxh4 2.Rd1 Rxd4 3.Qxd4 Qxb2 4.a5 Qa3 5.Na4 e5 6.Qh4 e4 7.Qh8+ Qf8
+/= (0.64) Depth: 9 00:00:00 1082kN
+/- (0.76) Depth: 13 00:00:02 23773kN

1.Rxf6 gxf6 2.Ne4 e5 3.Nxf6+ Nxf6 4.Qxe5+ Qe6 5.Qxe6+ fxe6 6.Bxf6 Rh7 7.Bxh4 Rc7 8.Rd1 Bd7 9.Rd4
+/- (1.15) Depth: 13 00:00:04 23773kN
+/- (1.10) Depth: 14 00:00:04 51626kN

1.Rd1 h3 2.Bc5 Qxd1+ 3.Kxd1 Nxc5 4.Bf3 Ncd7 5.Kc2 Ra7 6.Rd1 b6 7.g4 Bb7 8.g5
+- (1.55) Depth: 14 00:00:13 150mN
+- (2.25) Depth: 14 00:00:27 307mN
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
Top
User avatar
AdminX
Posts: 6363
Joined: Mon Mar 13, 2006 2:34 pm
Location: Acworth, GA

Re: Test Position---Not For Over Priced Commercials

Post by AdminX »

Given more time Crafty comes back to Rxf6.

Analysis by Crafty 22.5:

1.Ne4 Nxe4 2.Bxe4 hxg3 3.hxg3 Rh2 4.Bxg7 f5 5.Rh1
+/= (0.34) Depth: 7 00:00:00 323kN
= (0.19) Depth: 9 00:00:00 1082kN

1.gxh4 Rxh4 2.Rd1 Rxd4 3.Qxd4 Qxb2 4.a5 Qa3 5.Na4 e5 6.Qh4 e4 7.Qh8+ Qf8
+/= (0.64) Depth: 9 00:00:00 1082kN
+/- (0.76) Depth: 13 00:00:02 23773kN

1.Rxf6 gxf6 2.Ne4 e5 3.Nxf6+ Nxf6 4.Qxe5+ Qe6 5.Qxe6+ fxe6 6.Bxf6 Rh7 7.Bxh4 Rc7 8.Rd1 Bd7 9.Rd4
+/- (1.15) Depth: 13 00:00:04 23773kN
+/- (1.10) Depth: 14 00:00:04 51626kN

1.Rd1 Rh5 2.Rxf6 gxf6 3.Bc6 bxc6 4.Bxf6 Qxd1+ 5.Kxd1 Nxf6 6.Qxc6+ Kd8 7.Qxa8 hxg3 8.hxg3 Rh2 9.Qb8 Nd7 10.Qb4 f6
+- (1.55) Depth: 14 00:00:13 150mN
+- (2.60) Depth: 18 00:15:19 2147mN

1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7 5.Bc5 0-0-0 6.Bb6 Qc4 7.Qe7 Kb8 8.Qxd8+ Qc8 9.Nxf6 Qxd8 10.Bxd8 Be6 11.Kf2 b6 12.gxh4 Rxh4
+- (2.85) Depth: 18 00:19:36 2147mN
+- (3.99) Depth: 20 00:49:24 2147mN
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
Top
User avatar
Eelco de Groot
Posts: 4671
Joined: Sun Mar 12, 2006 2:40 am
Full name:   Eelco de Groot

Back to 1. Rd1, Build 137 and 103 of Toga 1.4 Checkov Beta 4

Post by Eelco de Groot »

Ovyron wrote:
Eelco de Groot wrote:Probably a real tactician does much better here, Deep Fritz 11 or Winfinder?
Winfinder doesn't find it in reasonable time. The problem may be that g4 also wins and it's not designed to find the "best win."
Thanks Ulysses, interesting! 1. g4 does not show up here in the PVs of Toga so maybe Winfinder is seeing something different after this move. If Winfinder does a lot of pruning after it finds a promising move, it is possible that she(?) overlooks another winning move. That is why I'm impressed with what I see of Rybka 3 in this regard, I think all Rybka versions do a lot of pruning for unpromising moves, but Rybka 3 seems to see much more of the winning sacrifices than Rybka 2.3.2a. Probably because a lot of Winfinder experiences where added to Rybka 3.

This is what I'm getting with the latest build of the single reply extensions, build 137 took a long time to complete 19 ply deep iteration for 1. Rd1 but 1. Rxf6 was rejected fairly quickly again at 19 plies.


r1b1k2r/1p1n1pp1/p2Qpn2/8/P2B3p/1qN3P1/1P2P1BP/R3KR2 w kq -

Engine: Toga 1.4 Checkov Beta 4/Toga Mara Beta 6 Build 137 (Athlon 2009 MHz, 256 MB)
by Fabien Letouzey, Thomas Gaksch, Taban Juma, Eelco de Groot

8/46 0:04 +1.47 1.Rd1 e5 2.Nd5 Nxd5 3.Bxd5 Rh6
4.Qxh6 Qxd5 5.Qg5 hxg3 6.hxg3 (2.577.216) 555

9/46 0:10 +0.68 1.Rd1 Rh5 2.gxh4 Rxh4 3.Qc7 Rh6
4.Rd2 Ng4 5.h3 Ngf6 6.Bxf6 Rxf6 (5.668.534) 561

9/46 0:14 +1.14 1.Rf3 Qc2 2.Rd1 Rh6 3.Bc5 Ng8 4.Be3 Rh5
5.g4 Qxb2 (8.130.111) 557

10/46 0:17 +1.02 1.Rf3 Qc2 2.Rd1 h3 3.Bc5 Ng8 4.Bf1 Qxb2
5.Rd2 Qb3 6.a5 (9.532.683) 555

10/52 0:21 +1.40 1.Rd1 Rh5 2.gxh4 Kd8 3.Bxf6+ gxf6
4.Rd2 Rxh4 5.Rxf6 Ke8 6.Be4 Rg4 7.a5 Rg1+
8.Kf2 (12.239.292) 560

11/60 0:48 +1.83 1.Rd1 Rh5 2.gxh4 a5 3.Bxf6 gxf6
4.Nb5 Rc5 5.Nc7+ Rxc7 6.Qxc7 f5 7.h5 Qb4+
8.Rd2 Qxa4 9.h6 Qh4+ 10.Rf2 Rb8 (28.424.077) 581

12/60 1:26 +1.89 1.Rd1 hxg3 2.Bxf6 gxf6 3.Ne4 Rh6
4.Rxf6 Rg6 5.Rxg6 fxg6 6.hxg3 a5
7.Rd2 Qa2 8.Ng5 Qb1+ 9.Rd1 (50.364.510) 582

13/60 3:05 +2.71 1.Rd1 hxg3 2.Rd3 Qc4 3.b3 Qxb3 4.Nd5 Qb1+
5.Rd1 Qxd1+ 6.Kxd1 Nxd5 7.Bxd5 exd5
8.Qxg3 Rh7 9.Bxg7 Nf8 10.Bxf8 Kxf8
11.Qd6+ Ke8 12.Qxd5 Be6 (108.417.356) 584

14/60 8:16 +2.19 1.Rd1 e5 2.Nd5 Nxd5 3.Bxd5 Rh6
4.Qxh6 Qxd1+ 5.Kxd1 gxh6 6.Be3 f6
7.gxh4 Ke7 8.h5 Kd6 9.Be4 Ke7
10.Bxh6 (298.496.371) 601

15/63 14:50 +1.99 1.Rd1 e5 2.Nd5 Nxd5 3.Bxd5 Rh6
4.Qxh6 Qxd1+ 5.Kxd1 gxh6 6.Be3 f6
7.gxh4 Ke7 8.h5 Kd6 9.Be4 Ke7
10.Bxh6 Nc5 (532.612.936) 598

15/66 21:55 +4.49 1.Rxf6 a5 2.Nb5 Qb4+ 3.Qxb4 axb4
4.Rxe6+ fxe6 5.Nc7+ Kf7 6.Nxa8 hxg3
7.hxg3 e5 8.Bf2 Nf6 9.Rc1 e4 10.Nb6 Bg4
11.Rc7+ Kg8 12.Rxb7 (797.408.917) 606

16/66 24:11 +4.56 1.Rxf6 a5 2.Nb5 Qb4+ 3.Qxb4 axb4
4.Rxe6+ fxe6 5.Nc7+ Kf7 6.Nxa8 hxg3
7.hxg3 e5 8.Bf2 Nf6 9.Rc1 Rh2 10.Bf3 Rh8
11.Nc7 Rg8 12.Kd2 Rd8+ 13.Ke3 (885.119.210) 609

17/67 30:36 +4.91 1.Rxf6 a5 2.Nb5 Qb4+ 3.Qxb4 axb4
4.Rxe6+ fxe6 5.Nc7+ Kf7 6.Nxa8 hxg3
7.hxg3 e5 8.Bg1 Nf6 9.Rc1 Bf5
10.Rc7+ Kg6 11.Bxb7 Be4 12.Nb6 Rf8
13.Bxe4+ Nxe4 14.a5 (1.126.827.076) 613

18/67 38:09 +4.90 1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6
4.Bxd7+ Bxd7 5.Bc5 O-O-O 6.Bb6 Qxb6
7.Qxb6 Bc6 8.Qe3 Rg6 9.Rc1 Rd4
10.Qa3 hxg3 11.Rxc6+ bxc6 12.Qf8+ Kd7
13.Nc5+ Kc7 14.Qxf7+ (1.415.943.612) 618

19/69 55:32 +2.74 1.Rxf6 gxf6 2.Ne4 e5 3.Bh3 Rh6
4.Bxd7+ Bxd7 5.Bc5 O-O-O 6.Bb6 Qc4
7.Qe7 Kb8 8.Qxd8+ Qc8 9.Qxc8+ Bxc8
10.Be3 Rh8 11.g4 f5 12.gxf5 a5 (2.061.603.977) 618

19/69 93:10 +3.26 1.Rd1 Rh5 2.Rd3 h3 3.Bf3 Rg5 4.Nb5 Rxb5
5.Bc5 Rxc5 6.Rxb3 a5 7.Rc3 Rxc3
8.bxc3 e5 9.e4 Kd8 10.Kf2 Ra6 11.Qd3 Ke7 (3.356.782.600) 600

20/82 516:55 +3.34 1.Rd1 h3 2.Bc5 Qxd1+ 3.Kxd1 Nxc5
4.Bf3 Ncd7 5.Kc1 Rg8 6.Rd1 g5 7.e4 e5
8.Bg4 Nxg4 9.Nd5 Nf8 10.Kb1 b5
11.axb5 axb5 12.Qc6+ Kd8 13.Nb6+ Nd7
14.Rxd7+ (17.925.531.508) 577


It may be however that the best follow-up for White after 1. Rxf6 gxf6 is not 2. Ne4, 2. Rd1 is also possible I saw in some other Builds where 2. Ne4 went down.

This is with earlier build 103 that does not do any of the fancy single reply extensions, does less Late Move Reductions but actually the main difference is in the experimental qsearch penalty for long searches that I don't trust much in build 103 but certainly not in later builds, but haven't gotten around in changing yet for all later versions.

After 1. Rxf6 gxf6 , 2.Ne4 goes down in multi-PV search.

[d]r1b1k2r/1p1n1p2/p2Qpp2/8/P2B3p/1qN3P1/1P2P1BP/R3K3 w kq -

Engine: Toga 1.4 Checkov Beta 4/Toga Mara Beta 6 Build 103 (Athlon 2009 MHz, 256 MB)
by Fabien Letouzey, Thomas Gaksch, Taban Juma, Eelco de Groot


14 1:37 +3.81 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7
5.Bc5 O-O-O 6.Bb6 Qc4 7.Qe7 Kb8
8.Qxd8+ Bc8 9.Nd6 Qc6 10.Ba7+ Ka8 (60.202.543) 619

14 2:14 +2.65 2.Rd1 Rh5 3.Rd3 Qxb2 4.Ne4 Qc1+
5.Rd1 Qxd1+ 6.Kxd1 Rd5 7.Nxf6+ Nxf6
8.Bxd5 Nxd5 9.e4 Ne7 10.Bf6 Nc6
11.gxh4 e5 12.Bxe5 Be6 13.Bf6 Bb3+
14.Kd2 Bxa4 (83.785.532) 619

14 1:53 +2.63 2.Nd5 exd5 3.Ra3 Qxa3 4.bxa3 Rg8
5.Qf4 hxg3 6.hxg3 Rg6 7.Bxd5 Rg5
8.e4 Rh5 9.g4 Ne5 10.Qxf6 Rh1+
11.Kd2 Rh2+ 12.Bf2 (70.448.594) 619

14 2:20 +2.56 2.Bd5 Qxb2 3.Rb1 Qxb1+ 4.Nxb1 hxg3
5.Qxg3 exd5 6.Nc3 Rh5 7.Qf3 Rh4
8.Nxd5 Rxd4 9.Qe3+ Kf8 10.Qxd4 Kg7
11.h4 Ne5 12.h5 Bf5 13.h6+ Kh8
14.Nxf6 (87.679.797) 619

14 2:30 +1.86 2.Ra3 Qxb2 3.Nd5 Qc1+ 4.Kf2 exd5
5.Re3+ Qxe3+ 6.Kxe3 hxg3 7.hxg3 Rh5
8.Bxd5 a5 9.Ke4 Rg5 10.e3 Rh5 11.Qf4 Ne5
12.Qxf6 Ra6 13.Qg7 Bf5+ 14.Kf4 (93.555.139) 619

14 2:57 +0.90 2.gxh4 Rxh4 3.Rd1 Rxd4 4.Rxd4 Qb6
5.Qxb6 Nxb6 6.h4 f5 7.h5 Nd7 8.a5 Ke7
9.h6 Nf6 10.Bf3 Rb8 11.Rh4 Nh7
12.Rd4 (109.782.600) 619

_______________________________________________________________

15 3:17 +3.36 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7
5.Bc5 O-O-O 6.Bb6 Qc4 7.Qe7 Kb8
8.Qxd8+ Qc8 9.Qxc8+ Kxc8 10.gxh4 Rxh4
11.Rc1+ Bc6 12.Rd1 Rh8 13.Nd6+ Kb8
14.Nxf7 (123.443.614) 629

15 3:30 +2.75 2.Rd1 Rh5 3.Rd3 Qxb2 4.Ne4 Qc1+
5.Rd1 Qxd1+ 6.Kxd1 Rd5 7.Nxf6+ Nxf6
8.Bxd5 Nxd5 9.e4 Ne7 10.Bf6 Nc6
11.gxh4 e5 12.Kd2 Bg4 13.Bxe5 Rd8 (131.376.628) 629

15 3:40 +2.68 2.Nd5 exd5 3.Ra3 Qxa3 4.bxa3 Rg8
5.Qf4 hxg3 6.hxg3 Rg6 7.Bxd5 Rg5
8.e4 Rh5 9.g4 Ne5 10.Qxf6 Rh1+
11.Kd2 Rh2+ 12.Kc3 Rh3+ 13.Kc2 (137.887.110) 629

15 3:47 +2.18 2.Bd5 Qxb2 3.Rb1 Qxb1+ 4.Nxb1 hxg3
5.Qxg3 exd5 6.Nc3 Rh5 7.Qf3 Rg5
8.Nxd5 Kf8 9.Nxf6 Nxf6 10.Qxf6 Rd5
11.e4 Rxd4 12.Qh8+ Ke7 13.Qxd4 Be6
14.Qb4+ Kf6 15.Qxb7 (142.722.297) 629

15 4:27 +2.03 2.Ra3 Qxb2 3.Nd5 Qc1+ 4.Kf2 exd5
5.Re3+ Qxe3+ 6.Kxe3 hxg3 7.hxg3 Rh5
8.Bxd5 Rf5 9.Kd2 Rxd5 10.Qxd5 Ne5
11.Bxe5 fxe5 12.Qxe5+ Kf8 13.Qh8+ Ke7
14.a5 Kd7 15.Qd4+ (168.414.456) 629

15 5:04 +0.72 2.gxh4 Rxh4 3.Ne4 Rxe4 4.Bxe4 e5
5.Bc3 Qe6 6.Qxe6+ fxe6 7.h4 f5 8.Bc2 Kf7
9.Rd1 Kg6 10.Rd6 e4 11.Rxe6+ Kh5
12.Re7 Kxh4 13.Rh7+ Kg5 14.Rg7+ Kf4
15.Bd2+ (191.505.322) 629

_______________________________________________________________

16 7:33 +3.11 2.Rd1 Rh5 3.Rd3 Qc2 4.g4 Re5 5.Bxe5 fxe5
6.Bxb7 Qc1+ 7.Nd1 Ra7 8.Rc3 Qxc3+
9.Nxc3 Bxb7 10.g5 Ra8 11.Qb4 Bd5
12.Qxh4 (289.239.103) 639

16 7:06 +2.85 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7
5.Bc5 O-O-O 6.Bb6 Qc4 7.Qe7 Kb8
8.Qxd8+ Qc8 9.Qxc8+ Bxc8 10.Be3 Rh8
11.Rd1 hxg3 12.hxg3 Rh1+ 13.Kd2 (271.882.717) 639

16 7:52 +2.69 2.Nd5 exd5 3.Ra3 Qxa3 4.bxa3 Rg8
5.Qf4 hxg3 6.hxg3 Rg6 7.Bxd5 Rg5
8.e4 Rh5 9.g4 Ne5 10.Qxf6 Rh1+
11.Kd2 Rh2+ 12.Kc3 Rh3+ 13.Kc2 Rh2+
14.Bf2 (301.106.472) 639

16 8:34 +2.54 2.g4 Rg8 3.Bf3 Rg5 4.Rd1 a5 5.Bxf6 Qxd1+
6.Kxd1 Nxf6 7.Qf4 Rg6 8.g5 Nd7 9.Nb5 e5
10.Qxh4 Kf8 11.h3 Raa6 12.Qh8+ Rg8 (328.202.320) 639

16 8:11 +2.30 2.Bd5 Qxb2 3.Rb1 Qxb1+ 4.Nxb1 hxg3
5.Qxg3 exd5 6.Nc3 Rh5 7.Qf3 Rg5
8.Nxd5 Kf8 9.Nxf6 Nxf6 10.Qxf6 Rd5
11.e4 Rxd4 12.Qh8+ Ke7 13.Qxd4 Kf8
14.h4 Kg8 15.h5 (313.777.951) 639

16 9:11 +2.06 2.Ra3 Qxb2 3.Nd5 Qc1+ 4.Kf2 exd5
5.Re3+ Qxe3+ 6.Kxe3 hxg3 7.hxg3 Rh5
8.Bxd5 Rf5 9.Kd2 Rxd5 10.Qxd5 Ne5
11.Bxe5 fxe5 12.Qxe5+ Kf8 13.Ke3 a5
14.Ke4 Kg8 15.Qe7 (352.412.800) 639

________________________________________________________________

17 9:32 +3.14 2.Rd1 Rh5 3.Rd3 Qc2 4.g4 Re5 5.Bxe5 fxe5
6.Bxb7 Qc1+ 7.Nd1 Ra7 8.Rc3 Qxc3+
9.Nxc3 Bxb7 10.g5 Kd8 11.e4 Ke8
12.b4 (365.981.896) 648

17 14:06 +2.83 2.Nd5 exd5 3.Ra3 Qxa3 4.bxa3 Rg8
5.Qf4 hxg3 6.hxg3 Rg6 7.Bxd5 Rg5
8.e4 Rh5 9.g4 Rh1+ 10.Kd2 Rh3 11.Qc7 Rh7
12.a5 Ne5 (548.338.678) 648

17 12:24 +2.79 2.Ne4 e5 3.Bh3 Rh6 4.Bxd7+ Bxd7
5.Bc5 O-O-O 6.Bb6 Qc4 7.Qe7 Kb8
8.Qxd8+ Qc8 9.g4 Qxd8 10.Bxd8 f5
11.Nf6 Be6 12.g5 (481.979.722) 648

17 15:29 +2.66 2.Bd5 Qxb2 3.Rb1 Qxb1+ 4.Nxb1 hxg3
5.Qxg3 exd5 6.Nc3 Rh5 7.Qf3 Rg5
8.Nxd5 Kf8 9.Nxf6 Nxf6 10.Qxf6 Rd5
11.e4 Rxd4 12.Qh8+ Ke7 13.Qxd4 Kf8
14.Qh8+ Ke7 15.h4 (601.278.921) 648

17 14:50 +2.65 2.g4 Rg8 3.Bf3 Rg5 4.Rd1 Qc2 5.Bxf6 Qxd1+
6.Kxd1 Nxf6 7.Qf4 Rg6 8.g5 Nd7
9.Qxh4 Rg8 10.Ne4 Ke7 11.Qh7 Rd8 (576.377.996) 648

17 16:49 +2.17 2.Ra3 Qxb2 3.Nd5 Qc1+ 4.Kf2 exd5
5.Re3+ Qxe3+ 6.Kxe3 hxg3 7.hxg3 Rh5
8.Bxd5 Rf5 9.a5 Rh5 10.g4 Rxd5
11.Qxd5 Ne5 12.Bxe5 fxe5 13.Qxe5+ Kd8
14.Kd4 Kd7 15.e4 (654.324.909) 648

Thanks for all the comparison material Ted and Swami :D !

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
Ovyron
Posts: 4562
Joined: Tue Jul 03, 2007 4:30 am

Re: Back to 1. Rd1, Build 137 and 103 of Toga 1.4 Checkov Be

Post by Ovyron »

Here's a Winfinder comparison.

After 1.Rxf6:

14.01 0:52 +0.90 1...gxf6 2.Ne4 e5 3.Bh3 Rh6 4.Nc5 Qb4+ 5.Bc3 f5 6.Qxh6 Qxc5 (2.097.521) 40

After 1.g4:

13.02 1:15 +1.61 1...e5 2.Rxf6 Nxf6 3.Nd5 Qxd5 4.Bxd5 exd4 5.Bc6+ Nd7 6.Bxd7+ Bxd7 7.Qxd4 O-O-O 8.Qxg7 (2.924.867) 39

Winfinder thinks that delaying Rxf6 is better!
Top

Return to “Computer Chess Club: General Topics”