The limits of "Just-mobility-evaluation"

[Edsel Apostol]

Hi Oliver,

Pawn Structure seems to become really valuable only in endgame. In the middlegame it was being compensated by mobility and piece activity. So when you add Pawn Structure to Olithink, try to have different scoring for opening/middlegame and endgame, then just combine this two scores through scaling by material on the board.

[OliverBr]

OliThink 5.0.8 has no pawn structure at all and has easily a rating of 2700 on ICC.

You have definitely done a great job on your engine, but from what I observed in our games I think you may not yet be at a reliable 2700 on ICC yet (unless you were to play only vs. humans, which seems to give all engines who do that a several hundred point pop). I would suggest playing a bunch of games with the non-rybka engines that regularly show up (Tinker, Arasan, Amateur, Now, LearningLemming, Crafty) to get a good feel for where your engine stacks up.

I have tried to limit Lemming's # of games with you because some people get very annoyed at playing the same engine over and over again (Lemming plays with auto-rematch) but would be happy to log in and play bunches of games if you would like.

-Sam

So you are the Author of Lemming?; nice. I have no problem when you play as often as you want. Rating is not so important;)

You are quite right, my engine ist quite weak against other engines. It performes much better against humans because of the uncommon style of player.

[OliverBr]

14.0-0-0 Nc5 15.Qe3 Nxd3+ 16.Rxd3 a5 17.Qe5 Qxe5 18.fxe5 Rf8 19.Rg1 f6 20.Kd2 fxe5
= (-0.19) Depth: 18 00:02:15 8456kN

Could someone in this group please explain how Rybka does search 18 plies with just 8456kN?
It just seems to double the nodes each ply. Incredible.

There its strength comes from, I suppose...

[Ovyron]

Could someone in this group please explain how Rybka does search 18 plies with just 8456kN?
It just seems to double the nodes each ply. Incredible.

Remember that Rybka uses some sort of obfuscation and doesn't report its nodes like other engines.

[Uri Blass]

Could someone in this group please explain how Rybka does search 18 plies with just 8456kN?
It just seems to double the nodes each ply. Incredible.

Remember that Rybka uses some sort of obfuscation and doesn't report its nodes like other engines.

Yes but Rybka depth is also not correct and you need to add 3 to it based on reverse engineering of Juri Osipov so it is practically depth 21.

The fact that rybka has branching factor of 2 is correct and she does it by aggresive reductions(the idea is to reduce depth when you expect fail low by what is called late move reductions).

The reason for branching factor of 2 is mainly agreesive reductions
Toga that is free source code also use reductions and has branching factor that is close to 2.

21: New game, 2008
rn1qk2r/pp3ppp/2p1p3/3p3n/3P1P1P/1PNB1P2/P1PQ1P2/R3K2R b KQkq - 0 1

Analysis by Toga II 1.4 beta5c:

1...Nh5-g3 2.f2xg3
+- (3.60) Depth: 1/2 00:00:00
1...Qd8-f6 2.Nc3-e2 c6-c5 3.c2-c3 Nb8-d7 4.f4-f5
µ (-0.77) Depth: 1/2 00:00:00
1...Qd8-f6 2.Nc3-e2 c6-c5 3.c2-c3 Nb8-d7 4.f4-f5
µ (-0.77) Depth: 6/14 00:00:00
1...Qd8-f6 2.Nc3-e2 c6-c5 3.c2-c3 Nb8-d7 4.f4-f5
µ (-0.77) Depth: 6/14 00:00:00
1...Qd8-f6 2.Nc3-e2 c6-c5 3.c2-c3 Nb8-d7 4.f4-f5
µ (-0.77) Depth: 6/14 00:00:00
1...Qd8-f6 2.Nc3-e2 c6-c5 3.c2-c3 Nb8-d7 4.f4-f5
µ (-0.77) Depth: 6/14 00:00:00
1...Qd8-f6 2.Nc3-e2 c6-c5 3.c2-c3 Nb8-d7 4.f4-f5
µ (-0.77) Depth: 6/14 00:00:00
1...Qd8-f6 2.Nc3-e2 c6-c5 3.c2-c3 Nb8-d7 4.f4-f5
µ (-0.77) Depth: 6/14 00:00:00
1...Qd8-f6 2.Nc3-e2 c6-c5 3.c2-c3 Nb8-d7 4.f4-f5
µ (-0.77) Depth: 6/14 00:00:00 9kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.f4-f5 c5xd4 4.f5xe6 Qf6xf3 5.e6xf7+ Ke8xf7 6.0-0-0 Nb8-c6
³ (-0.65) Depth: 7/22 00:00:00 24kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.f4-f5 Nb8-c6 4.f5xe6 f7xe6 5.Bd3-b5 c5xd4 6.Bb5xc6+ b7xc6 7.Ne2xd4 Qf6-e5+ 8.Qd2-e3
µ (-0.76) Depth: 8/22 00:00:00 65kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.Bd3-b5+ Nb8-d7 4.0-0-0 Ra8-c8 5.Qd2-e3 c5xd4 6.Rd1xd4 Qf6-f5 7.Bb5xd7+ Ke8xd7
³ (-0.66) Depth: 9/26 00:00:00 136kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.Bd3-b5+ Nb8-d7 4.0-0-0 Ra8-c8 5.Qd2-e3 c5xd4 6.Rd1xd4 Qf6-f5 7.Bb5xd7+ Ke8xd7 8.Qe3-e4 Qf5xe4 9.f3xe4
³ (-0.69) Depth: 10/26 00:00:00 212kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.Bd3-b5+ Nb8-d7 4.0-0-0 Ra8-c8 5.Qd2-e3 c5xd4 6.Rd1xd4 Qf6-f5 7.Bb5xd7+ Ke8xd7 8.Qe3-e4 Qf5xe4 9.f3xe4
³ (-0.69) Depth: 11/26 00:00:00 373kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.Bd3-b5+ Nb8-d7 4.0-0-0 Ra8-c8 5.Bb5xd7+ Ke8xd7 6.Kc1-b1 c5xd4 7.Ne2xd4 Nh5xf4 8.Qd2-a5 a7-a6 9.Rh1-g1
³ (-0.56) Depth: 12/34 00:00:01 814kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.Bd3-b5+ Nb8-c6 4.Bb5xc6+ b7xc6 5.Qd2-c3 c5-c4 6.b3xc4 0-0 7.Qc3-e3 g7-g6 8.c4xd5 c6xd5 9.Ra1-b1 Ra8-b8
³ (-0.52) Depth: 13/34 00:00:03 2132kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.Bd3-b5+ Nb8-c6 4.Bb5xc6+ b7xc6 5.Qd2-c3 c5-c4 6.Qc3-e3 Qf6-g6 7.Qe3-e5 Qg6-g2 8.Rh1-g1 Qg2xf3 9.Qe5-c7 0-0
³ (-0.52) Depth: 14/34 00:00:05 4191kN
1...Qd8-f6 2.Nc3-e2 Nb8-d7 3.0-0-0 0-0-0 4.Kc1-b1 h7-h6 5.c2-c3 Kc8-b8 6.Kb1-b2 Rd8-g8 7.Kb2-a1
³ (-0.58) Depth: 15/40 00:00:16 12301kN
1...Qd8-f6 2.Nc3-e2 c6-c5 3.Bd3-b5+ Nb8-c6 4.0-0-0 0-0 5.Bb5xc6 b7xc6 6.Rh1-g1 c5xd4 7.Qd2xd4 Qf6xd4 8.Ne2xd4 c6-c5 9.Nd4-e2 Ra8-b8 10.Kc1-b2 Nh5-f6
³ (-0.60) Depth: 16/43 00:00:31 23367kN
1...Qd8-f6 2.Nc3-e2 g7-g6 3.0-0-0 Nb8-d7 4.c2-c3 0-0-0 5.Kc1-b2 Kc8-b8 6.f4-f5 e6-e5 7.Qd2-g5 Rd8-c8 8.Kb2-a1 Qf6-d6 9.f5xg6 h7xg6
³ (-0.56) Depth: 17/43 00:01:11 52915kN

(, 30.01.2008)

Uri

[Dann Corbit]

14.0-0-0 Nc5 15.Qe3 Nxd3+ 16.Rxd3 a5 17.Qe5 Qxe5 18.fxe5 Rf8 19.Rg1 f6 20.Kd2 fxe5
= (-0.19) Depth: 18 00:02:15 8456kN

Could someone in this group please explain how Rybka does search 18 plies with just 8456kN?
It just seems to double the nodes each ply. Incredible.

There its strength comes from, I suppose...

Rybka counts nodes differently than other programs.

However, a branching factor of 2 is not unusual today.

Null move reductions + Late move reductions will get you into that ballpark.

Programs with a branching factor of about 2:
Shredder
Rybka
Chess Tiger
Wildcat
{lots more, I'm sure}

Open source programs with a branching factor of about 2:
Toga II
Glaurung
{lots more, I'm sure}

Toga II and Glaurung output for these two positions:
1br2rk1/1pqb1ppp/p3pn2/8/1P6/P1N1PN1P/1B3PP1/1QRR2K1 w - - bm Ne4; id "WAC.152";
1k1r4/pp1r1pp1/4n1p1/2R5/2Pp1qP1/3P2QP/P4PB1/1R4K1 w - - bm Bxb7; id "WAC.238";

Code: Select all

Analysis from E:\tprobs.epd   
1/29/2008 9:37:45 PM Level: 180 Seconds
Analyzing engine: Glaurung-32bit

1) Ne4; id "WAC.152";   
    Searching move: Nc3-e4
    Best move (Glaurung-32bit): Nc3-e4
    identical moves! Found in: 00:00
      2	00:00	         271	1.084	+0.15	Nc3b5 Qc7b6 Bb2xf6 Rc8xc1 Rd1xc1 Bd7xb5
      2	00:00	         482	1.812	+1.79	Nc3d5 Nf6xd5 Rc1xc7 Bb8xc7
      2	00:00	         844	3.172	+2.15	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6
      3	00:00	       1.129	4.244	+2.15	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6
      4	00:00	       1.608	6.045	+2.15	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6
      5	00:00	       3.392	12.028	+2.01	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Bb8a7
      5	00:00	       6.996	24.808	+2.25	Nc3d5 Nf6xd5 Rc1xc7 Bb8xc7 Nf3g5
      6	00:00	       9.094	30.619	+1.82	Nc3d5 Nf6xd5 Rc1xc7 Bb8xc7 Nf3g5 f7f5
      6	00:00	      13.117	44.164	+2.64	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Bb8a7 Nf3g5
      7	00:00	      22.433	71.670	+2.70	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Bf6xd8 Rf8xd8
      8	00:00	      37.554	104.316	+2.56	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Bf6xd8 Rf8xd8 Qb1b2
      9	00:00	      68.688	162.767	+2.56	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Bf6xd8 Rf8xd8 Qb1b2 Qc7b6 Rd1xd8+ Qb6xd8
     10	00:00	     126.226	230.760	+2.67	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Rd1xd8 Rf8xd8 Rc1c4 h7h5 Nf3e5
     11	00:00	     275.572	309.283	+2.89	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8d8 Rd1xd8+ Rc8xd8 Rc1c4 Rd8d5 Rc4g4+ Kg8f8 b4b5 Rd5xb5 Qb1xh7 Bc6xf3 g2xf3
     12	00:01	     573.954	363.492	+2.50	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8d8 Bf6xd8 Rc8xd8 Rc1c4 Qc7e7 Rc4c5 Rd8xd1+ Qb1xd1 Qe7f6
     13	00:03	   1.224.813	395.867	+2.71	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8d8 Bf6xd8 Rc8xd8 Rc1c4 f7f6 Rd1xd8+ Qc7xd8 Rc4g4+ Kg8h8 Qb1c2 Bc6xf3 g2xf3
     14	00:06	   2.721.602	412.739	+2.62	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8d8 Bf6xd8 Rc8xd8 Rc1c4 f7f5 Rd1xd8+ Qc7xd8 Rc4d4 Qd8e7 Qb1c2 Bb8c7
     15	00:15	   6.164.437	405.848	+2.71	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8d8 Bf6xd8 Rc8xd8 Rd1xd8+ Qc7xd8 Nf3d4 Bc6a4 Qb1e4 b7b6 Qe4b7 a6a5 b4xa5 b6xa5
     16	00:34	  14.547.473	418.788	+3.01	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 b4b5 a6xb5 Bb2xf6 Rf8d8 Nf3g5 Rd8xd1+ Rc1xd1 Qc7h2+ Kg1f1 Qh2xg2+ Kf1e1 Bc6e4 Qb1xe4 Qg2xe4 Ng5xe4 Bb8h2 f2f4 Kg8f8
     17	00:57	  24.586.765	426.757	+3.07	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 b4b5 a6xb5 Bb2xf6 Rf8d8 Nf3g5 Rd8xd1+ Rc1xd1 Qc7h2+ Kg1f1 Qh2xg2+ Kf1e1 Bc6e4 Qb1xe4 Qg2xe4 Ng5xe4 Bb8c7 Rd1b1 Bc7a5+ Ke1d1 b5b4 a3xb4
     18	01:51	  48.050.006	430.609	+3.37	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 b4b5 a6xb5 Bb2xf6 Rf8d8 Nf3g5 Rd8xd1+ Rc1xd1 Qc7h2+ Kg1f1 Qh2xg2+ Kf1e1 Bc6e4 Qb1xe4 Qg2xe4 Ng5xe4 Bb8c7 Rd1b1 Bc7d8 Rb1xb5 Bd8xf6 Ne4xf6+ Kg8g7
   1/29/2008 9:40:48 PM, Time for this analysis: 00:03:00, Rated time: 00:00

2) Bxb7; id "WAC.238";  
    Searching move: Bg2xb7
    Best move (Glaurung-32bit): Bg2xb7
    identical moves! Found in: 00:00
      2	00:00	         408	0	+0.89	Rc5a5 Qf4xg3 f2xg3
      2	00:00	         546	0	+0.90	Rc5e5 Rd8h8
      2	00:00	         942	0	+1.37	Rc5b5 b7b6
      3	00:00	       1.300	0	+1.10	Rc5b5 b7b6 Rb5e5 Qf4xg3 f2xg3
      4	00:00	       2.216	0	+1.10	Rc5b5 b7b6 Rb5e5 Qf4xg3 f2xg3
      5	00:00	       5.093	0	+1.03	Rc5b5 b7b6 Qg3f3 Qf4xf3 Bg2xf3
      5	00:00	       8.166	0	+3.82	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8xb7 Qg3g2+ Rd8d5 Qg2xd5+ Kb7b8 Rc5b5+ Kb8c7
      6	00:00	      22.731	733.258	+3.79	Bg2xb7 a7a6 Bb7c8+ Kb8a7 Rc5c6 Qf4xg3+ f2xg3 Rd8xc8 Rc6xc8
      7	00:00	      33.540	713.617	+3.79	Bg2xb7 a7a6 Bb7c8+ Kb8a7 Rc5c6 Qf4xg3+ f2xg3 Rd8xc8 Rc6xc8
      8	00:00	      60.740	557.247	+3.82	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Rc5b5 Qf4xg3+ f2xg3 Ne6g5 Kg1g2 Rd8e8
      9	00:00	      83.712	597.942	+4.12	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 a7a6 Rc7xf7 Nf4xd3 Rf7xg7
     10	00:00	     123.200	526.495	+4.37	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xh3+ Kg1f1 Nh3f4 Rb7xa7+ Ka8b8 Ra7b7+ Kb8a8 Rc7xf7 Nf4xd3 Rf7xg7
     11	00:00	     295.072	525.039	+5.25	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8xb7 Qg3g2+ Rd8d5 Qg2xd5+ Kb7b8 Rc5b5+ Kb8c8 Qd5b7+ Kc8d8 Rb5d5+ Kd8e8 Rd5d7 Qf4f6 Qb7c8+ Ne6d8 Rd7xa7 Ke8f8
     12	00:01	     569.842	528.610	+5.21	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 f7f6 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7d7 Nf4xh3+ Kg1f1
     13	00:01	     892.349	533.701	+5.31	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Rd8f8 Rb7xa7+ Ka8b8 Rc7xf7 Nf4xd3 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rg7f7
     14	00:02	   1.512.592	529.063	+5.50	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Rd8f8 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 h3h4 Nf4e2+ Kg1f1 Ne2f4 Rb7xf7 Rf8xf7 Ra7xf7 Nf4xd3 Rf7xg7
     15	00:05	   3.062.000	523.956	+6.15	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7d7 Rd8xd7 Rg7xd7 Nd3f4 Rd7xd4 Nf4xh3+ Kg1g2
     16	00:10	   5.452.445	512.399	+5.92	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7d7 Rd8xd7 Rg7xd7 Nd3f4 Rd7d8+ Ka8b7 Rd8xd4 Nf4xh3+ Kg1g2
     17	00:20	  10.676.415	518.398	+6.14	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xh3+ Kg1f1 Nh3f4 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Rg7b7+ Kb8c8 Rb7b3 Rd8d7 Ra7xd7 Kc8xd7 a2a4
     18	00:41	  20.920.663	509.837	+5.98	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xh3+ Kg1f1 Nh3f4 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Rg7b7+ Kb8c8 Rb7b3 Rd8d7 Ra7xd7 Kc8xd7 a2a4 Kd7c6
     19	01:31	  48.405.503	530.076	+5.98	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Rg7b7+ Kb8c8 Rb7c7+ Kc8b8 Ra7b7+ Kb8a8 c4c5 Nd3xc5 Rb7b6 Nc5e4 Rb6xg6 d4d3 h3h4
   1/29/2008 9:43:48 PM, Time for this analysis: 00:03:00, Rated time: 00:00

Code: Select all

Analysis from E:\tprobs.epd   
1/29/2008 9:31:20 PM Level: 180 Seconds
Analyzing engine: TogaIISMP

1) Ne4; id "WAC.152";   
    Searching move: Nc3-e4
    Best move (TogaIISMP): Nc3-e4
    identical moves! Found in: 00:00
     2/8	00:00	         499	0	+1.57	Nc3d5 Nf6xd5 Rc1xc7 Bb8xc7
     2/11	00:00	         941	0	+1.84	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6
     3/11	00:00	       1.256	0	+1.84	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6
     4/16	00:00	       3.147	0	+1.62	Nc3e4 Bd7c6 Bb2xf6 Bc6xe4 Rc1xc7 Be4xb1 Rc7xb7
     5/16	00:00	       5.125	0	+1.67	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Qc7b6 Nf3d4
     6/18	00:00	       9.863	0	+2.09	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Bb8a7 Nf3e5
     7/24	00:00	      24.436	0	+2.37	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 h7h6 Rc1c5 Bb8a7
     8/29	00:00	      43.017	0	+2.17	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8e8 Bf6e5 Qc7b6
     9/32	00:00	      82.439	0	+2.49	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8e8 Rc1c5 h7h6 Qb1b3 Bc6d5 Rc5xc7 Bd5xb3
    10/32	00:00	     135.906	0	+2.53	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8e8 Rc1c5 h7h6 Qb1b3 Bb8a7 Bf6e5
    11/32	00:00	     246.993	0	+2.69	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rf8e8 Rc1c5 h7h6 Nf3e5 Bb8a7 Rc5c4 Qc7b8 Ne5xc6 Rc8xc6 Rc4xc6 b7xc6
    12/34	00:00	     554.134	0	+3.21	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Bf6xd8 Rf8xd8 Rd1xd8+ Qc7xd8 Nf3d4 Bc6d7 Qb1e4 e6e5 Nd4f5
    13/34	00:01	   1.091.510	790.000	+3.29	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 h7h6 Rc1c4 Rf8d8 Rd1xd8+ Rc8xd8 Rc4g4+ Kg8f8 Bf6xd8 Qc7xd8 Nf3d4 h6h5 Nd4xc6 b7xc6 Rg4d4 Bb8d6 Qb1h7
    14/40	00:02	   2.164.962	800.000	+3.36	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Bf6xd8 Rf8xd8 Rd1xd8+ Qc7xd8 Nf3d4 Bc6d5 e3e4 Bb8a7 e4xd5 Ba7xd4 d5xe6 f7xe6 Qb1a2 Qd8d7
    15/48	00:05	   4.223.199	816.000	+3.31	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Bf6xd8 Rf8xd8 Rd1xd8+ Qc7xd8 Nf3d4 Bc6d5 e3e4 Bb8a7 e4xd5 Ba7xd4 d5xe6 f7xe6 Qb1a2 Qd8d7 Qa2d2 Qd7d6 Qd2g5+ Kg8f7
    16/48	00:09	   7.542.828	821.111	+3.36	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Bf6xd8 Rf8xd8 Rd1xd8+ Qc7xd8 Nf3d4 Bc6d5 e3e4 Bb8a7 e4xd5 Ba7xd4 d5xe6 f7xe6 Qb1a2 Qd8d6 Qa2c4 Qd6d7
    17/48	00:17	  14.430.653	825.882	+3.31	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 Bb2xf6 Rc8d8 Bf6xd8 Rf8xd8 Rd1xd8+ Qc7xd8 Nf3d4 Bc6d5 e3e4 Bb8a7 e4xd5 Ba7xd4 d5xe6 f7xe6 Qb1a2 Kg8g7 Qa2d2 Qd8d6 Qd2g5+ Kg7f7
    18/58	00:44	  36.388.310	820.909	+3.83	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 b4b5 a6xb5 Bb2xf6 Rf8d8 Nf3g5 Rd8xd1+ Rc1xd1 Qc7h2+ Kg1f1 Qh2xg2+ Kf1e1 Bc6e4 Qb1xe4 Qg2xe4 Ng5xe4 Bb8h2 f2f4 h7h5 Rd1d2 Rc8c1+ Ke1e2 Kg8h7
    19/58	01:26	  70.338.379	814.651	+3.90	Nc3e4 Bd7c6 Ne4xf6+ g7xf6 b4b5 a6xb5 Bb2xf6 Rf8d8 Nf3g5 Rd8xd1+ Rc1xd1 Qc7h2+ Kg1f1 Qh2xg2+ Kf1e1 Bc6e4 Qb1xe4 Qg2xe4 Ng5xe4 h7h6 Rd1d7 Bb8c7 Ke1e2 Kg8f8 Ke2f3 e6e5 Ne4c5
   1/29/2008 9:34:24 PM, Time for this analysis: 00:03:00, Rated time: 00:00

2) Bxb7; id "WAC.238";  
    Searching move: Bg2xb7
    Best move (TogaIISMP): Bg2xb7
    identical moves! Found in: 00:00
     2/11	00:00	       1.258	0	+1.29	Rc5b5 b7b6
     3/11	00:00	       2.017	0	+1.49	Rc5b5 b7b6 Qg3h4
     4/11	00:00	       2.743	0	+0.91	Rc5b5 b7b6 Qg3h4 g6g5
     5/12	00:00	       4.751	0	+0.61	Rc5b5 b7b6 Qg3h4 Qf4h6 Qh4xh6 g7xh6
     6/14	00:00	      10.163	0	+0.57	Rc5b5 b7b6 Qg3f3 Qf4xf3 Bg2xf3 Rd8e8
     6/25	00:00	      33.488	0	+3.51	Bg2xb7 a7a6 Bb7c6+ Kb8a7 Qg3xf4 Ne6xf4 Bc6xd7 Rd8xd7
     7/26	00:00	      50.745	0	+4.34	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rb7xf7 Nf4xd3 Rc5a5 Ka8b8 Ra5xa7 Nd3c5 Rf7xg7
     8/26	00:00	      63.010	0	+4.21	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rb7xf7 Nf4xd3 Rc5a5 Ka8b8 Ra5xa7 Nd3c5 Rf7xg7 d4d3 Rg7e7
     9/26	00:00	     106.913	0	+3.62	Bg2xb7 a7a6 Bb7xa6+ Kb8a7 Qg3xf4 Ne6xf4 Ba6c8 Ka7a8 Bc8xd7 Rd8xd7 Rb1b3 Nf4xh3+ Kg1f1
    10/39	00:00	     228.469	0	+4.87	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xh3+ Kg1f1 Nh3g5 Rb7xa7+ Ka8b8 f2f4 Ng5f3 Ra7b7+ Kb8a8 Rc7xf7 Nf3h2+ Kf1g1 Nh2xg4 Rf7xg7
    11/47	00:00	     408.726	0	+5.31	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xh3+ Kg1f1 Nh3g5 Rb7xa7+ Ka8b8 f2f4 Ng5e6 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Ra7b7+ Kb8c8 c4c5 Rd8h8
    12/47	00:01	     982.340	850.000	+5.18	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Rg7b7+ Kb8c8 Rb7e7 Kc8b8 Ra7b7+ Kb8a8 a2a4 Rd8h8 Rb7d7 Ka8b8 Rd7xd4 Rh8xh3
    13/47	00:01	   1.458.232	840.826	+5.26	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7e7 Rd8f8 Rg7xg6 Nd3xf2 Rg6a6+ Ka8b8 Ra6b6+ Kb8c8 Kg1g2
    14/50	00:03	   2.547.384	843.333	+5.41	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7e7 Rd8f8 Rg7xg6 Nd3xf2 h3h4 d4d3 Rg6d6 Nf2h3+ Kg1h2 Nh3f4 h4h5 Ka8b8
    15/56	00:06	   5.254.891	836.667	+5.31	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7e7 Nd3c5 Kg1f1 d4d3 Kf1e1 d3d2+ Ke1d1 Nc5d3 Kd1xd2 Nd3xf2+ Kd2c3
    16/56	00:11	   9.790.610	844.545	+5.46	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7e7 Nd3c5 Kg1f1 d4d3 Kf1e1 d3d2+ Ke1d1 Nc5d3 Kd1xd2 Nd3xf2+ Kd2c3 Nf2xh3 Rg7xg6 Nh3f4 Rg6a6+ Ka8b8 Ra6b6+ Kb8c8
    17/66	00:21	  17.603.797	823.991	+5.56	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7e7 Nd3c5 Kg1f1 Nc5d3 Rg7xg6 Rd8f8 Rg6a6+ Ka8b8 Ra6b6+ Kb8c8 Re7e2 Kc8d7
    18/66	00:37	  31.157.927	839.490	+5.63	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7e7 Nd3c5 Kg1f1 d4d3 Kf1e1 d3d2+ Ke1d1 Ka8b8 h3h4 Nc5d3 Kd1xd2 Nd3xf2+ Kd2e2 Nf2d3 Re7b7+ Kb8a8
    19/74	01:27	  73.091.171	836.730	+5.98	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xd3 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Ra7b7+ Kb8a8 Rb7e7 Nd3c5 Kg1f1 d4d3 Kf1e1 d3d2+ Ke1d1 Ka8b8 h3h4 Nc5d3 Kd1xd2 Nd3xf2+ Kd2e2 Nf2h3 Re7b7+ Kb8a8
    20/74	02:55	 147.528.496	843.644	+6.43	Bg2xb7 Rd7xb7 Rb1xb7+ Kb8a8 Qg3xf4 Ne6xf4 Rc5c7 Nf4xh3+ Kg1f1 Nh3f4 Rb7xa7+ Ka8b8 Rc7b7+ Kb8c8 Rb7xf7 Kc8b8 Rf7b7+ Kb8c8 Rb7xg7 Kc8b8 Rg7b7+ Kb8c8 Rb7f7 Kc8b8 Ra7b7+ Kb8a8 Rb7b5 Nf4e6 Rb5a5+ Ka8b8 Rf7f6 Ne6c7 Rf6xg6 Kb8b7 c4c5 Nc7d5 a2a4 Rd8c8 Ra5b5+ Kb7c7
   1/29/2008 9:37:27 PM, Time for this analysis: 00:03:02, Rated time: 00:00

[Uri Blass]

14.0-0-0 Nc5 15.Qe3 Nxd3+ 16.Rxd3 a5 17.Qe5 Qxe5 18.fxe5 Rf8 19.Rg1 f6 20.Kd2 fxe5
= (-0.19) Depth: 18 00:02:15 8456kN

Could someone in this group please explain how Rybka does search 18 plies with just 8456kN?
It just seems to double the nodes each ply. Incredible.

There its strength comes from, I suppose...

I can add that the reductions are not reductions that cost nothing in the same depth relative to no reductions so the difference in strength from the reductions is smaller than the difference that is suggested by the difference in depths that you see.

Uri

[FrancoisK]

I agree.
But i think that there is a lot coming also from the evaluation.
A few month ago I compared Glaurung and BugChess2 (which has a significantly more complicated evaluation) over a thousand actual game positions after removing all "fancy" stuff like null move, LMR, futility in order to be as close as possible to the bare alpha beta skeleton - i even changed my QS pruning to match Glaurung's.
In these conditions, with a move ordering scheme roughly equivalent, Glaurung's branch factor was clearly much better. It seems to be most of comes from a much better (though or because it was simpler) and stable evaluation (in addition to being twice or 3x faster). Cf Tord's post on evaluation. Or there's a big bug lurking somewhere

François

[Marimar]

I believe that it is a common mistake of humans to overevaluate pawn structure and one of the reasons that rybka is better is the fact that rybka has smaller scores for bad pawn structure relative to most programs.

Uri

This sounds reasonable. OliThink 5.0.8 has no pawn structure at all and has easily a rating of 2700 on ICC.
But on the other hand, most games OliThink loses against other engines (It has not lost against any human, GM, IM or whatever yet, looks like a "human killer") were lost because in endgames the other engines eat all the pawns that have been completely unstructured. So did Rybka in that very game.

We have to put some things in context about humans overevaluating pawn structure. Humans "evaluate" positions which occur actually in games, or which can be reached from them by a short sequence of good moves.
On the other hand, computers evaluate leaf positions of a deep search tree, most of them look very strange to human eyes and can not be correctly assesed with traditional human chess wisdom.
So I guess both strong human and comp chess players have correct "evaluactions" when it comes to pawn structure, in the respective context of the positions in which each one is applying them.

Now I have some questions for you. Does OliThink really evaluate only mobility? I mean, does it have a material evaluation? Anyway it sounds to be a very interesting approach.
Now, I guess that in order to improve you need to add some sort of pawn structure evaluation. But critical to strong positional play is the search policy. Which reductions and extensions do you use?

José.