preprocessing by houdini1.5

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

Moderators: hgm, Rebel, chrisw

Uri Blass
Posts: 10309
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

preprocessing by houdini1.5

Post by Uri Blass »

It seems that the evaluation of houdini is dependent on the root position

before Qxe7 houdini evaluates the position as at least 1.8 for white at depths 20-24

After Qxe7 Rxe7 houdini does not get the 1.8 number at depth 20-25

I wonder if this preprocessing is a bug or some intended behaviour and if houdini2.0 shows the same behaviour.

New game - Houdini 1.5 w32, 90'+30"/40+30'+30"
[D]6k1/pp2qp1p/3p1Qp1/3Pr2P/2p5/8/PP2B2P/3K1R2 w - - 0 1

Analysis by Houdini 1.5 w32:

26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6
+- (1.86) Depth: 4/13 00:00:00
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6
+- (1.86) Depth: 5/13 00:00:00
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6 b7-b5 30.Kd1-d2 f7-f5
+- (2.09) Depth: 6/13 00:00:00 4kN
26.Qf6xe7
+- (2.19) Depth: 7/17 00:00:00 6kN
26.Qf6xe7
+- (2.41) Depth: 7/17 00:00:00 8kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 g6-g5 28.Rf6xd6 b7-b5 29.a2-a4 b5xa4 30.Be2xc4
+- (2.41) Depth: 7/18 00:00:00 10kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.h5xg6 h7xg6 29.Be2xc4 Kg8-g7 30.Rf6-f3 Rd7-c7
+- (1.82) Depth: 8/21 00:00:01 25kN
26.Qf6xe7
+- (1.92) Depth: 9/23 00:00:01 34kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 g6-g5 30.Rf4-f5 f7-f6 31.Bc4-f1 Rd7-e7 32.Kd1-c2 g5-g4
+- (1.81) Depth: 9/23 00:00:01 40kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-e7 30.h5xg6 h7xg6 31.Kd1-c2 f7-f5 32.Bc4-f1 Re7-e5 33.Rf4-b4
+- (1.74) Depth: 10/23 00:00:01 90kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-e7 30.Kd1-d2 g6xh5 31.Rf4-f5 f7-f6 32.Bc4-d3 Re7-e5 33.Kd2-c3 a7-a6
+- (1.80) Depth: 11/26 00:00:01 137kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-e7 30.h5xg6 h7xg6 31.Rf1-e1 Re7xe1+ 32.Kd1xe1 Kg7-f6 33.Ke1-d2 Kf6-e5 34.Kd2-e3 f7-f5 35.h2-h4
+- (1.89) Depth: 12/26 00:00:01 197kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-e7 30.h5xg6 h7xg6 31.Rf1-e1 Re7xe1+ 32.Kd1xe1 Kg7-f6 33.Ke1-d2 Kf6-e5 34.Kd2-e3 f7-f5 35.h2-h4
+- (1.89) Depth: 13/26 00:00:01 276kN
26.Qf6xe7 Re5xe7
+- (1.79) Depth: 14/28 00:00:01 458kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-c7 30.b2-b3 Rc7-e7 31.h5xg6 h7xg6 32.Rf1-e1 Re7xe1+ 33.Kd1xe1 Kg7-f6 34.Ke1-d2 Kf6-e5 35.Kd2-e3 f7-f5 36.Bc4-f1
+- (1.80) Depth: 14/28 00:00:01 602kN
26.Qf6xe7 Re5xe7 27.Be2xc4 Re7-e4 28.Bc4-b3 Kg8-f8 29.h5xg6 h7xg6 30.Rf1-f2 Re4-h4 31.Rf2-e2 b7-b5 32.Kd1-c1 Rh4-h3 33.Bb3-c2 Kf8-g7 34.Kc1-d2 f7-f5 35.Bc2-d3 a7-a6 36.Re2-e7+ Kg7-h6
+- (1.73) Depth: 15/33 00:00:02 1603kN
26.Qf6xe7 Re5xe7
+- (1.62) Depth: 16/33 00:00:03 2746kN
26.Qf6xe7 Re5xe7 27.Be2xc4 Re7-e4 28.Bc4-b3 Re4-h4 29.h5xg6 h7xg6 30.Rf1-f6 Kg8-f8 31.Rf6-f2 a7-a5 32.Kd1-d2 f7-f5 33.Kd2-e3 Kf8-f7 34.Rf2-e2 a5-a4 35.Bb3-c2 Kf7-f6 36.Re2-g2 g6-g5 37.Rg2-e2 Rh4-h3+ 38.Ke3-d4 Rh3-h4+ 39.Kd4-c3
+- (1.61) Depth: 16/40 00:00:06 4828kN
26.Qf6xe7 Re5xe7 27.Be2xc4 Re7-e4 28.Bc4-b3 Re4-h4 29.h5xg6 h7xg6 30.Rf1-f2 Kg8-g7 31.Kd1-d2 f7-f5 32.Rf2-e2 Kg7-f7 33.Kd2-e3 b7-b5 34.Bb3-c2 Rh4-h3+ 35.Ke3-f4 Rh3-h4+ 36.Kf4-f3 Rh4-d4 37.Re2-e6 Rd4xd5
+- (1.66) Depth: 17/45 00:00:09 7333kN
26.Qf6xe7
+- (1.76) Depth: 18/45 00:00:13 11169kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 a7-a6 31.Kd1-c2 b7-b5 32.Bc4-f1 Rh4-d4 33.Rf2-d2 Rd4-f4 34.Bf1-h3 f7-f5 35.Rd2-e2 Kg8-f7 36.Bh3-g2 Rf4-h4
+- (1.78) Depth: 18/45 00:00:15 13401kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.a2-a4 Kg7-f6 33.Kd1-c2 Rh4-h3 34.Kc2-d2 Kf6-g5 35.Re2-g2+ Kg5-f6 36.Rg2-f2+ Kf6-g7 37.Rf2-e2 b7-b6 38.Bc4-d3 Rh3-h5 39.Kd2-c3 Rh5xd5 40.Re2-e7 Rd5-e5
+- (1.77) Depth: 19/45 00:00:18 15661kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 Kg7-f6 32.Kd1-c2 Kf6-g5 33.Re2-g2+ Rh4-g4 34.Rg2xg4+ Kg5xg4 35.Bc4-b5 f7-f5 36.Kc2-d3 g6-g5 37.Bb5-d7 Kg4-f4 38.Bd7-e6 Kf4-e5
+- (1.81) Depth: 20/47 00:00:26 23004kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.a2-a4 Kg7-f6 33.Kd1-d2 Rh4-h3 34.Kd2-c2 b7-b6 35.Re2-f2+ Kf6-g7 36.Bc4-f1 Rh3-h5 37.Rf2-d2 Rh5-f5
+- (1.90) Depth: 21/47 00:00:38 34341kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Kg8-g7 30.Rf1-f2 f7-f5 31.Rf2-e2 Re4xe2 32.Kd1xe2 g6-g5 33.a2-a4 Kg7-f6 34.Ke2-f3 Kf6-e5 35.Bc4-d3 Ke5-f6 36.a4-a5 Kf6-e5 37.Bd3-c2 Ke5-f6 38.h2-h3 Kf6-e5 39.b3-b4 b7-b6 40.a5xb6 a7xb6 41.Bc2-b1 Ke5-f6
+- (1.83) Depth: 22/51 00:00:55 49604kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Kg8-g7 30.Rf1-f2 f7-f5 31.Rf2-e2 Re4xe2 32.Kd1xe2 g6-g5 33.a2-a4 Kg7-f6 34.Ke2-f3 Kf6-e5 35.Bc4-d3 Ke5-f6 36.a4-a5 Kf6-e5 37.Bd3-c2 Ke5-f6 38.h2-h3 b7-b6 39.a5xb6 a7xb6 40.b3-b4 Kf6-e5 41.Bc2-b1 Ke5-f6
+- (1.83) Depth: 23/51 00:01:26 77577kN
26.Qf6xe7
+- (1.93) Depth: 24/51 00:02:36 141948kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Kg8-g7 30.Rf1-f2 Re4-e5 31.Rf2-e2 Kg7-f6 32.Kd1-d2 a7-a6 33.a2-a4 Re5-h5 34.Kd2-c3 a6-a5 35.Kc3-d4 Rh5-h4+ 36.Kd4-e3 Kf6-g5 37.Re2-g2+ Kg5-h6 38.Rg2-f2 Kh6-g7 39.Rf2-e2 f7-f5 40.Re2-g2 Rh4-e4+ 41.Ke3-f3
+- (1.93) Depth: 24/51 00:02:51 156842kN

(, 22.05.2012)

New game - Houdini 1.5 w32, 90'+30"/40+30'+30"
[D]6k1/pp2rp1p/3p2p1/3P3P/2p5/8/PP2B2P/3K1R2 w - - 0 1

Analysis by Houdini 1.5 w32:

27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6
+- (1.60) Depth: 4/11 00:00:00
27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6 b7-b5 30.Kd1-d2 f7-f5
+- (1.79) Depth: 5/12 00:00:00
27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6 b7-b5 30.Kd1-d2 f7-f5
+- (1.79) Depth: 6/14 00:00:00 4kN
27.h5xg6 h7xg6 28.Rf1-f6 Re7-d7 29.Be2xc4 Kg8-g7 30.Rf6-f1 Rd7-c7
+- (1.49) Depth: 7/16 00:00:00 12kN
27.Be2xc4 f7-f5 28.Bc4-d3 Kg8-g7 29.Kd1-d2 Re7-e8 30.h2-h4 a7-a6 31.h5xg6 h7xg6
+- (1.58) Depth: 7/16 00:00:00 13kN
27.Be2xc4 f7-f5 28.Bc4-d3 Kg8-g7 29.Kd1-d2 Re7-e8 30.h2-h4 a7-a6 31.h5xg6 h7xg6
+- (1.58) Depth: 8/16 00:00:00 17kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-e7 30.h5xg6 h7xg6 31.Kd1-d2 Re7-e5 32.h2-h4
+- (1.60) Depth: 9/21 00:00:01 45kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-c7 30.h5xg6 h7xg6 31.Kd1-e2 f7-f5 32.h2-h4 Kg7-f6 33.b2-b3 Kf6-e5
+- (1.58) Depth: 10/21 00:00:01 61kN
27.Rf1-f6 Re7-d7
+- (1.49) Depth: 11/21 00:00:01 95kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-c7 30.h5xg6 h7xg6 31.h2-h4 f7-f5 32.Kd1-e2 Kg7-f6 33.Ke2-f3 Kf6-e5 34.b2-b3
+- (1.55) Depth: 11/22 00:00:01 144kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-c7 30.b2-b3 a7-a6 31.Kd1-d2 b7-b5 32.Bc4-d3 f7-f5 33.Kd2-e3 Rc7-e7+ 34.Ke3-f4
+- (1.55) Depth: 12/24 00:00:01 203kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-c7 30.b2-b3 Rc7-e7 31.h5xg6 h7xg6 32.Rf1-e1 Re7xe1+ 33.Kd1xe1 Kg7-f6 34.Ke1-e2 Kf6-e5 35.Ke2-e3
+- (1.56) Depth: 13/24 00:00:01 325kN
27.Rf1-f6 Re7-d7
+- (1.47) Depth: 14/33 00:00:01 685kN
27.Rf1-f6 Kg8-f8 28.h5xg6 h7xg6 29.Be2xc4 Re7-e4 30.b2-b3 Kf8-e7 31.Rf6-f2 Re4-e5 32.Kd1-d2 f7-f5 33.h2-h4 Ke7-f6 34.Rf2-f4 a7-a6 35.Kd2-d3
+- (1.45) Depth: 14/33 00:00:01 1001kN
27.Rf1-f6 Kg8-f8 28.h5xg6 h7xg6 29.Be2xc4 Re7-e4 30.b2-b3 Kf8-e7 31.Rf6-f2 Re4-e5 32.Rf2-e2 f7-f5 33.h2-h4 Ke7-f6 34.Re2xe5 Kf6xe5 35.Kd1-e2 Ke5-e4 36.b3-b4
+- (1.45) Depth: 15/33 00:00:01 1229kN
27.Rf1-f6 Kg8-f8 28.h5xg6 h7xg6 29.Be2xc4 Re7-e4 30.b2-b3 Re4-h4 31.Rf6-f2 Kf8-g7 32.Rf2-e2 a7-a5 33.a2-a4 Rh4-h3 34.Re2-e7 b7-b6 35.Re7-d7 Rh3xh2 36.Rd7xd6
+- (1.49) Depth: 16/33 00:00:02 1782kN
27.Rf1-f6 Kg8-f8 28.h5xg6 h7xg6 29.Be2xc4 Re7-e4 30.b2-b3 Kf8-e7 31.Rf6-f2 Re4-e5 32.Kd1-c2 f7-f5 33.h2-h4 Ke7-f6 34.Kc2-d3 a7-a6 35.a2-a4 Re5-e4 36.Rf2-e2
+- (1.43) Depth: 17/35 00:00:03 2858kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f6 Kg8-f8 31.Rf6-f2 a7-a5 32.Rf2-e2 Rh4-h8 33.a2-a3 Rh8-h3 34.Kd1-d2 Kf8-g7 35.Bc4-d3 Rh3-h5 36.Kd2-c3 Rh5xd5 37.Re2-e7 Rd5-c5+ 38.Kc3-d4
+- (1.52) Depth: 17/35 00:00:05 4493kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a6 32.a2-a4 Rh4-h3 33.Re2-e7 b7-b5 34.a4xb5 a6xb5 35.Bc4xb5 Rh3xb3 36.Bb5-e2 Rb3-b2 37.h2-h4
+- (1.51) Depth: 18/35 00:00:06 5703kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 Rh4-h3 32.a2-a4 Kg7-f6 33.Kd1-c2 a7-a6 34.Bc4-d3 Rh3-h5 35.Bd3-e4 Rh5-h3 36.a4-a5 Kf6-g5
+- (1.57) Depth: 19/38 00:00:10 8819kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.a2-a4 Rh4-h3 33.Kd1-c2 Kg7-f6 34.Re2-f2+ Kf6-g7 35.Bc4-d3 Rh3-h5 36.Kc2-c3 Rh5xd5 37.Bd3-c4 Rd5-f5 38.Rf2xf5 g6xf5
+- (1.60) Depth: 20/44 00:00:17 14416kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.Kd1-c2 Rh4-h3 33.a2-a3 Kg7-f6 34.Bc4-d3 Rh3-h5 35.Kc2-c3 Rh5xd5 36.Bd3-e4 Rd5-b5 37.Kc3-c4 Rb5-b6 38.Be4-d5
+- (1.62) Depth: 21/44 00:00:28 24619kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.Kd1-c2 Rh4-h3 33.a2-a3 Kg7-f6 34.a3-a4 Rh3-h4 35.Kc2-c3 Rh4-h5 36.Kc3-d3 Rh5-h3+ 37.Kd3-d4 Rh3-h4+ 38.Kd4-d3 Rh4-h3+ 39.Kd3-d4
+- (1.64) Depth: 22/48 00:00:49 44230kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.a2-a4 a7-a5 32.Kd1-d2 Rh4-h3 33.Bc4-d3 Rh3-h4 34.Kd2-c3 Rh4-h5 35.Bd3-f1 Rh5-g5 36.Kc3-d4 Rg5-g4+ 37.Kd4-e3 Rg4-g5 38.Bf1-c4 Rg5-h5 39.Rf2-e2 Kg7-f6 40.Ke3-f3 Rh5-f5+ 41.Kf3-e4 Rf5-h5 42.Ke4-d4 Rh5-h4+
+- (1.64) Depth: 23/54 00:01:05 59321kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.a2-a4 a7-a5 32.Kd1-d2 Rh4-h3 33.Bc4-d3 Rh3-h4 34.Kd2-c3 Rh4-h5 35.Bd3-f1 Rh5-g5 36.Kc3-d4 Rg5-g4+ 37.Kd4-e3 Rg4-g5 38.Ke3-e4 Rg5-e5+ 39.Ke4-d4 f7-f5 40.Rf2-f4 Re5-e1
+- (1.65) Depth: 24/54 00:01:52 103252kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.a2-a4 Rh4-h3 32.Rf2-e2 Kg7-f6 33.Kd1-c2 a7-a5 34.Bc4-d3 Rh3-h5 35.Kc2-c3 b7-b6 36.Re2-f2+ Kf6-e7 37.Bd3-c4 Rh5-h3+ 38.Kc3-d4 Rh3-h4+ 39.Kd4-e3 f7-f5 40.Rf2-e2 Rh4-e4+ 41.Ke3-f3
+- (1.72) Depth: 25/54 00:03:46 210655kN

(, 22.05.2012)
Top
User avatar
Houdini
Posts: 1471
Joined: Tue Mar 16, 2010 12:00 am

Re: preprocessing by houdini1.5

Post by Houdini »

Uri, you have a keen eye.
The small difference in evaluation is due to Houdini's rescaling of the output to produce a score in "pawns", which is done based on the material value of the root position.
Pawns get relatively more valuable near the end game, so the same material advantage in the end game is worth less "pawns" than in the middle game.

Robert
Top
User avatar
Strelkaman
Posts: 83
Joined: Tue May 15, 2012 1:38 pm

Re: preprocessing by houdini1.5

Post by Strelkaman »

With Houdini 2.0c pro Uri,Houdini evaluates as best moves 1.hxg6 | Evaluation: 2.00 | Depth: 24 | Possible continuation: 1. hxg6 hxg6 2. Qxe7 Rxe7 3. Bxc4 Re4 4. b3 Kg7 5. Re1 Rxe1+ 6. Kxe1 g5 7. Ke2 f5 8. Kf3 Kf6 9. Be2 Ke5 10. Bd3 b6 11. h3 a5 12. Bb1 Kf6 13. a3 Ke5 14. Bd3 f4 15. Be4 Kd4 16. b4 axb4 17. axb4 and 2.Qxe7 | Evaluation: 2.00 | Depth: 24 | Possible continuation:
1. Qxe7 Rxe7 2. hxg6 hxg6 3. Bxc4 Re4 4. b3 Kg7 5. Re1 Rxe1+ 6. Kxe1 g5 7. Ke2 f5 8. Kf3 Kf6 9. Be2 Ke5 10. Bd3 b6 11. h3 a5 12. Bb1 Kf6 13. a3 Ke5 14. Bd3 f4 15. Be4 Kd4 16. b4 axb4 17. axb4 with equal rating as you see (2.00)

Now if you follow the 1.hxg6 hxg6 2.Qxe7 Rxe7 you see that the 2.00 rating drops to 1.69-1.71 and 1.80(depth 24) for both proposals by Houdini

My explanation is that Houdini has separate evaluations with and without gaviota tablebases or maby drops by 20points the winning for white due to white king safety
Strelka is the best engine in the world!!
Top
Hood
Posts: 657
Joined: Mon Feb 08, 2010 12:52 pm
Location: Polska, Warszawa

Re: preprocessing by houdini1.5

Post by Hood »

Does Houdini Pro has the same engine as Houdini Standard. I men chess engine in SP cathegories.
Rgds Hood
Polish National tragedy in Smoleńsk. President and all delegation murdered or killed.
Cui bono ?

There are not bugs free programs.
There are programs with undiscovered bugs.




Ashes to ashes dust to dust. Alleluia.
Top
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: preprocessing by houdini1.5

Post by bob »

Uri Blass wrote:It seems that the evaluation of houdini is dependent on the root position

before Qxe7 houdini evaluates the position as at least 1.8 for white at depths 20-24

After Qxe7 Rxe7 houdini does not get the 1.8 number at depth 20-25

I wonder if this preprocessing is a bug or some intended behaviour and if houdini2.0 shows the same behaviour.

New game - Houdini 1.5 w32, 90'+30"/40+30'+30"
[D]6k1/pp2qp1p/3p1Qp1/3Pr2P/2p5/8/PP2B2P/3K1R2 w - - 0 1

Analysis by Houdini 1.5 w32:

26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6
+- (1.86) Depth: 4/13 00:00:00
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6
+- (1.86) Depth: 5/13 00:00:00
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6 b7-b5 30.Kd1-d2 f7-f5
+- (2.09) Depth: 6/13 00:00:00 4kN
26.Qf6xe7
+- (2.19) Depth: 7/17 00:00:00 6kN
26.Qf6xe7
+- (2.41) Depth: 7/17 00:00:00 8kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 g6-g5 28.Rf6xd6 b7-b5 29.a2-a4 b5xa4 30.Be2xc4
+- (2.41) Depth: 7/18 00:00:00 10kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.h5xg6 h7xg6 29.Be2xc4 Kg8-g7 30.Rf6-f3 Rd7-c7
+- (1.82) Depth: 8/21 00:00:01 25kN
26.Qf6xe7
+- (1.92) Depth: 9/23 00:00:01 34kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 g6-g5 30.Rf4-f5 f7-f6 31.Bc4-f1 Rd7-e7 32.Kd1-c2 g5-g4
+- (1.81) Depth: 9/23 00:00:01 40kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-e7 30.h5xg6 h7xg6 31.Kd1-c2 f7-f5 32.Bc4-f1 Re7-e5 33.Rf4-b4
+- (1.74) Depth: 10/23 00:00:01 90kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-e7 30.Kd1-d2 g6xh5 31.Rf4-f5 f7-f6 32.Bc4-d3 Re7-e5 33.Kd2-c3 a7-a6
+- (1.80) Depth: 11/26 00:00:01 137kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-e7 30.h5xg6 h7xg6 31.Rf1-e1 Re7xe1+ 32.Kd1xe1 Kg7-f6 33.Ke1-d2 Kf6-e5 34.Kd2-e3 f7-f5 35.h2-h4
+- (1.89) Depth: 12/26 00:00:01 197kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-e7 30.h5xg6 h7xg6 31.Rf1-e1 Re7xe1+ 32.Kd1xe1 Kg7-f6 33.Ke1-d2 Kf6-e5 34.Kd2-e3 f7-f5 35.h2-h4
+- (1.89) Depth: 13/26 00:00:01 276kN
26.Qf6xe7 Re5xe7
+- (1.79) Depth: 14/28 00:00:01 458kN
26.Qf6xe7 Re5xe7 27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-c7 30.b2-b3 Rc7-e7 31.h5xg6 h7xg6 32.Rf1-e1 Re7xe1+ 33.Kd1xe1 Kg7-f6 34.Ke1-d2 Kf6-e5 35.Kd2-e3 f7-f5 36.Bc4-f1
+- (1.80) Depth: 14/28 00:00:01 602kN
26.Qf6xe7 Re5xe7 27.Be2xc4 Re7-e4 28.Bc4-b3 Kg8-f8 29.h5xg6 h7xg6 30.Rf1-f2 Re4-h4 31.Rf2-e2 b7-b5 32.Kd1-c1 Rh4-h3 33.Bb3-c2 Kf8-g7 34.Kc1-d2 f7-f5 35.Bc2-d3 a7-a6 36.Re2-e7+ Kg7-h6
+- (1.73) Depth: 15/33 00:00:02 1603kN
26.Qf6xe7 Re5xe7
+- (1.62) Depth: 16/33 00:00:03 2746kN
26.Qf6xe7 Re5xe7 27.Be2xc4 Re7-e4 28.Bc4-b3 Re4-h4 29.h5xg6 h7xg6 30.Rf1-f6 Kg8-f8 31.Rf6-f2 a7-a5 32.Kd1-d2 f7-f5 33.Kd2-e3 Kf8-f7 34.Rf2-e2 a5-a4 35.Bb3-c2 Kf7-f6 36.Re2-g2 g6-g5 37.Rg2-e2 Rh4-h3+ 38.Ke3-d4 Rh3-h4+ 39.Kd4-c3
+- (1.61) Depth: 16/40 00:00:06 4828kN
26.Qf6xe7 Re5xe7 27.Be2xc4 Re7-e4 28.Bc4-b3 Re4-h4 29.h5xg6 h7xg6 30.Rf1-f2 Kg8-g7 31.Kd1-d2 f7-f5 32.Rf2-e2 Kg7-f7 33.Kd2-e3 b7-b5 34.Bb3-c2 Rh4-h3+ 35.Ke3-f4 Rh3-h4+ 36.Kf4-f3 Rh4-d4 37.Re2-e6 Rd4xd5
+- (1.66) Depth: 17/45 00:00:09 7333kN
26.Qf6xe7
+- (1.76) Depth: 18/45 00:00:13 11169kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 a7-a6 31.Kd1-c2 b7-b5 32.Bc4-f1 Rh4-d4 33.Rf2-d2 Rd4-f4 34.Bf1-h3 f7-f5 35.Rd2-e2 Kg8-f7 36.Bh3-g2 Rf4-h4
+- (1.78) Depth: 18/45 00:00:15 13401kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.a2-a4 Kg7-f6 33.Kd1-c2 Rh4-h3 34.Kc2-d2 Kf6-g5 35.Re2-g2+ Kg5-f6 36.Rg2-f2+ Kf6-g7 37.Rf2-e2 b7-b6 38.Bc4-d3 Rh3-h5 39.Kd2-c3 Rh5xd5 40.Re2-e7 Rd5-e5
+- (1.77) Depth: 19/45 00:00:18 15661kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 Kg7-f6 32.Kd1-c2 Kf6-g5 33.Re2-g2+ Rh4-g4 34.Rg2xg4+ Kg5xg4 35.Bc4-b5 f7-f5 36.Kc2-d3 g6-g5 37.Bb5-d7 Kg4-f4 38.Bd7-e6 Kf4-e5
+- (1.81) Depth: 20/47 00:00:26 23004kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.a2-a4 Kg7-f6 33.Kd1-d2 Rh4-h3 34.Kd2-c2 b7-b6 35.Re2-f2+ Kf6-g7 36.Bc4-f1 Rh3-h5 37.Rf2-d2 Rh5-f5
+- (1.90) Depth: 21/47 00:00:38 34341kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Kg8-g7 30.Rf1-f2 f7-f5 31.Rf2-e2 Re4xe2 32.Kd1xe2 g6-g5 33.a2-a4 Kg7-f6 34.Ke2-f3 Kf6-e5 35.Bc4-d3 Ke5-f6 36.a4-a5 Kf6-e5 37.Bd3-c2 Ke5-f6 38.h2-h3 Kf6-e5 39.b3-b4 b7-b6 40.a5xb6 a7xb6 41.Bc2-b1 Ke5-f6
+- (1.83) Depth: 22/51 00:00:55 49604kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Kg8-g7 30.Rf1-f2 f7-f5 31.Rf2-e2 Re4xe2 32.Kd1xe2 g6-g5 33.a2-a4 Kg7-f6 34.Ke2-f3 Kf6-e5 35.Bc4-d3 Ke5-f6 36.a4-a5 Kf6-e5 37.Bd3-c2 Ke5-f6 38.h2-h3 b7-b6 39.a5xb6 a7xb6 40.b3-b4 Kf6-e5 41.Bc2-b1 Ke5-f6
+- (1.83) Depth: 23/51 00:01:26 77577kN
26.Qf6xe7
+- (1.93) Depth: 24/51 00:02:36 141948kN
26.Qf6xe7 Re5xe7 27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Kg8-g7 30.Rf1-f2 Re4-e5 31.Rf2-e2 Kg7-f6 32.Kd1-d2 a7-a6 33.a2-a4 Re5-h5 34.Kd2-c3 a6-a5 35.Kc3-d4 Rh5-h4+ 36.Kd4-e3 Kf6-g5 37.Re2-g2+ Kg5-h6 38.Rg2-f2 Kh6-g7 39.Rf2-e2 f7-f5 40.Re2-g2 Rh4-e4+ 41.Ke3-f3
+- (1.93) Depth: 24/51 00:02:51 156842kN

(, 22.05.2012)

New game - Houdini 1.5 w32, 90'+30"/40+30'+30"
[D]6k1/pp2rp1p/3p2p1/3P3P/2p5/8/PP2B2P/3K1R2 w - - 0 1

Analysis by Houdini 1.5 w32:

27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6
+- (1.60) Depth: 4/11 00:00:00
27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6 b7-b5 30.Kd1-d2 f7-f5
+- (1.79) Depth: 5/12 00:00:00
27.h5xg6 h7xg6 28.Rf1-f6 Kg8-g7 29.Rf6xd6 b7-b5 30.Kd1-d2 f7-f5
+- (1.79) Depth: 6/14 00:00:00 4kN
27.h5xg6 h7xg6 28.Rf1-f6 Re7-d7 29.Be2xc4 Kg8-g7 30.Rf6-f1 Rd7-c7
+- (1.49) Depth: 7/16 00:00:00 12kN
27.Be2xc4 f7-f5 28.Bc4-d3 Kg8-g7 29.Kd1-d2 Re7-e8 30.h2-h4 a7-a6 31.h5xg6 h7xg6
+- (1.58) Depth: 7/16 00:00:00 13kN
27.Be2xc4 f7-f5 28.Bc4-d3 Kg8-g7 29.Kd1-d2 Re7-e8 30.h2-h4 a7-a6 31.h5xg6 h7xg6
+- (1.58) Depth: 8/16 00:00:00 17kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-e7 30.h5xg6 h7xg6 31.Kd1-d2 Re7-e5 32.h2-h4
+- (1.60) Depth: 9/21 00:00:01 45kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-c7 30.h5xg6 h7xg6 31.Kd1-e2 f7-f5 32.h2-h4 Kg7-f6 33.b2-b3 Kf6-e5
+- (1.58) Depth: 10/21 00:00:01 61kN
27.Rf1-f6 Re7-d7
+- (1.49) Depth: 11/21 00:00:01 95kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f4 Rd7-c7 30.h5xg6 h7xg6 31.h2-h4 f7-f5 32.Kd1-e2 Kg7-f6 33.Ke2-f3 Kf6-e5 34.b2-b3
+- (1.55) Depth: 11/22 00:00:01 144kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-c7 30.b2-b3 a7-a6 31.Kd1-d2 b7-b5 32.Bc4-d3 f7-f5 33.Kd2-e3 Rc7-e7+ 34.Ke3-f4
+- (1.55) Depth: 12/24 00:00:01 203kN
27.Rf1-f6 Re7-d7 28.Be2xc4 Kg8-g7 29.Rf6-f1 Rd7-c7 30.b2-b3 Rc7-e7 31.h5xg6 h7xg6 32.Rf1-e1 Re7xe1+ 33.Kd1xe1 Kg7-f6 34.Ke1-e2 Kf6-e5 35.Ke2-e3
+- (1.56) Depth: 13/24 00:00:01 325kN
27.Rf1-f6 Re7-d7
+- (1.47) Depth: 14/33 00:00:01 685kN
27.Rf1-f6 Kg8-f8 28.h5xg6 h7xg6 29.Be2xc4 Re7-e4 30.b2-b3 Kf8-e7 31.Rf6-f2 Re4-e5 32.Kd1-d2 f7-f5 33.h2-h4 Ke7-f6 34.Rf2-f4 a7-a6 35.Kd2-d3
+- (1.45) Depth: 14/33 00:00:01 1001kN
27.Rf1-f6 Kg8-f8 28.h5xg6 h7xg6 29.Be2xc4 Re7-e4 30.b2-b3 Kf8-e7 31.Rf6-f2 Re4-e5 32.Rf2-e2 f7-f5 33.h2-h4 Ke7-f6 34.Re2xe5 Kf6xe5 35.Kd1-e2 Ke5-e4 36.b3-b4
+- (1.45) Depth: 15/33 00:00:01 1229kN
27.Rf1-f6 Kg8-f8 28.h5xg6 h7xg6 29.Be2xc4 Re7-e4 30.b2-b3 Re4-h4 31.Rf6-f2 Kf8-g7 32.Rf2-e2 a7-a5 33.a2-a4 Rh4-h3 34.Re2-e7 b7-b6 35.Re7-d7 Rh3xh2 36.Rd7xd6
+- (1.49) Depth: 16/33 00:00:02 1782kN
27.Rf1-f6 Kg8-f8 28.h5xg6 h7xg6 29.Be2xc4 Re7-e4 30.b2-b3 Kf8-e7 31.Rf6-f2 Re4-e5 32.Kd1-c2 f7-f5 33.h2-h4 Ke7-f6 34.Kc2-d3 a7-a6 35.a2-a4 Re5-e4 36.Rf2-e2
+- (1.43) Depth: 17/35 00:00:03 2858kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f6 Kg8-f8 31.Rf6-f2 a7-a5 32.Rf2-e2 Rh4-h8 33.a2-a3 Rh8-h3 34.Kd1-d2 Kf8-g7 35.Bc4-d3 Rh3-h5 36.Kd2-c3 Rh5xd5 37.Re2-e7 Rd5-c5+ 38.Kc3-d4
+- (1.52) Depth: 17/35 00:00:05 4493kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a6 32.a2-a4 Rh4-h3 33.Re2-e7 b7-b5 34.a4xb5 a6xb5 35.Bc4xb5 Rh3xb3 36.Bb5-e2 Rb3-b2 37.h2-h4
+- (1.51) Depth: 18/35 00:00:06 5703kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 Rh4-h3 32.a2-a4 Kg7-f6 33.Kd1-c2 a7-a6 34.Bc4-d3 Rh3-h5 35.Bd3-e4 Rh5-h3 36.a4-a5 Kf6-g5
+- (1.57) Depth: 19/38 00:00:10 8819kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.a2-a4 Rh4-h3 33.Kd1-c2 Kg7-f6 34.Re2-f2+ Kf6-g7 35.Bc4-d3 Rh3-h5 36.Kc2-c3 Rh5xd5 37.Bd3-c4 Rd5-f5 38.Rf2xf5 g6xf5
+- (1.60) Depth: 20/44 00:00:17 14416kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.Kd1-c2 Rh4-h3 33.a2-a3 Kg7-f6 34.Bc4-d3 Rh3-h5 35.Kc2-c3 Rh5xd5 36.Bd3-e4 Rd5-b5 37.Kc3-c4 Rb5-b6 38.Be4-d5
+- (1.62) Depth: 21/44 00:00:28 24619kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.Rf2-e2 a7-a5 32.Kd1-c2 Rh4-h3 33.a2-a3 Kg7-f6 34.a3-a4 Rh3-h4 35.Kc2-c3 Rh4-h5 36.Kc3-d3 Rh5-h3+ 37.Kd3-d4 Rh3-h4+ 38.Kd4-d3 Rh4-h3+ 39.Kd3-d4
+- (1.64) Depth: 22/48 00:00:49 44230kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.a2-a4 a7-a5 32.Kd1-d2 Rh4-h3 33.Bc4-d3 Rh3-h4 34.Kd2-c3 Rh4-h5 35.Bd3-f1 Rh5-g5 36.Kc3-d4 Rg5-g4+ 37.Kd4-e3 Rg4-g5 38.Bf1-c4 Rg5-h5 39.Rf2-e2 Kg7-f6 40.Ke3-f3 Rh5-f5+ 41.Kf3-e4 Rf5-h5 42.Ke4-d4 Rh5-h4+
+- (1.64) Depth: 23/54 00:01:05 59321kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.a2-a4 a7-a5 32.Kd1-d2 Rh4-h3 33.Bc4-d3 Rh3-h4 34.Kd2-c3 Rh4-h5 35.Bd3-f1 Rh5-g5 36.Kc3-d4 Rg5-g4+ 37.Kd4-e3 Rg4-g5 38.Ke3-e4 Rg5-e5+ 39.Ke4-d4 f7-f5 40.Rf2-f4 Re5-e1
+- (1.65) Depth: 24/54 00:01:52 103252kN
27.h5xg6 h7xg6 28.Be2xc4 Re7-e4 29.b2-b3 Re4-h4 30.Rf1-f2 Kg8-g7 31.a2-a4 Rh4-h3 32.Rf2-e2 Kg7-f6 33.Kd1-c2 a7-a5 34.Bc4-d3 Rh3-h5 35.Kc2-c3 b7-b6 36.Re2-f2+ Kf6-e7 37.Bd3-c4 Rh5-h3+ 38.Kc3-d4 Rh3-h4+ 39.Kd4-e3 f7-f5 40.Rf2-e2 Rh4-e4+ 41.Ke3-f3
+- (1.72) Depth: 25/54 00:03:46 210655kN

(, 22.05.2012)
It can also be something else... Trading queens certainly changes king safety. And that can cause a different alpha/beta window to be used after the queens are off at the root position. And different alpha/beta window values can then lead to different reduction/pruning/evaluation/search behavior...
Top
User avatar
marcelk
Posts: 348
Joined: Sat Feb 27, 2010 12:21 am

Re: preprocessing by houdini1.5

Post by marcelk »

Houdini wrote:Uri, you have a keen eye.
The small difference in evaluation is due to Houdini's rescaling of the output to produce a score in "pawns", which is done based on the material value of the root position.
Pawns get relatively more valuable near the end game, so the same material advantage in the end game is worth less "pawns" than in the middle game.

Robert
The obvious question I think is: have you considered applying this concept in the leaves instead of the root? Hypothetically, if it could be done without additional CPU cycles, do you think that would be helpful?
Top
User avatar
Houdini
Posts: 1471
Joined: Tue Mar 16, 2010 12:00 am

Re: preprocessing by houdini1.5

Post by Houdini »

marcelk wrote:
Houdini wrote:Uri, you have a keen eye.
The small difference in evaluation is due to Houdini's rescaling of the output to produce a score in "pawns", which is done based on the material value of the root position.
Pawns get relatively more valuable near the end game, so the same material advantage in the end game is worth less "pawns" than in the middle game.

Robert
The obvious question I think is: have you considered applying this concept in the leaves instead of the root? Hypothetically, if it could be done without additional CPU cycles, do you think that would be helpful?
Doing the rescaling at the leaves would hurt the strength, it would prevent the engine from searching for simplification (swap material) when it has a material lead.
Top
mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 9:17 pm

Re: preprocessing by houdini1.5

Post by mcostalba »

Houdini wrote: Doing the rescaling at the leaves would hurt the strength, it would prevent the engine from searching for simplification (swap material) when it has a material lead.
So this is not just "rescaling of the output", this affects search behaviour, so it is a functionality change. (Nice idea btw).
Top
BubbaTough
Posts: 1154
Joined: Fri Jun 23, 2006 5:18 am

Re: preprocessing by houdini1.5

Post by BubbaTough »

mcostalba wrote:
Houdini wrote: Doing the rescaling at the leaves would hurt the strength, it would prevent the engine from searching for simplification (swap material) when it has a material lead.
So this is not just "rescaling of the output", this affects search behaviour, so it is a functionality change. (Nice idea btw).

Where do you get that? it sounds to me like he is just changing the output, assumably because the adjusted values look prettier to the human eye but are not actually valid for comparing positions.

-Sam
Top

Return to “Computer Chess Club: General Topics”