Computer Chess Club Archives


Search

Terms

Messages

Subject: Another tactical horizon effect in search reproduced

Author: Vincent Diepeveen

Date: 17:33:52 05/06/01

Go up one level in this thread


On May 06, 2001 at 02:28:14, Uri Blass wrote:

>I gave Deep Fritz to analyze similiar number of nodes to Deeper blue and Deep
>Fritz seems to be clearly better in tactics.

Hi Uri,

Here another tactical horizon effect of Deep Blue and that
is when the first search line shows up about the f5 move at move 28.

This line shows up in deep blue at move 27 of black at 10 ply.

black timeleft=27:46.40.00
 - = - r r = k =   d3xe4    20    Bd6-c5
 = o = n = o = -   Nf1-e3   21    Ra8-d8
 - q o = - = - o   Nh2-f1   22    g5-g4
 o - = - o - b b   h3xg4    23    Nf6xg4
 - = - = O = - =   f2-f3    24    Ng4xe3
 O O = - N O O -   Nf1xe3   25    Bc5-e7
 - B O = R = B =   Kg1-h1   26    Be7-g5
 R - Q - = - = K   Re1-e2   27    ...
white timeleft=27:46.40.00
black to move               type 'help' for command overview

clean
Clearing all Hashtables...
anal
Analysis mode is on
Process 1: engineflags=256 msk=1
process 0: engineflags = 256 denktime=10000000 maxtime=10000000
White can win here 0 worth of material
Black can win here 0 worth of material
00:00 0 0 7 (0) 1 -0.15 Bg5xe3 Qc1xe3
00:00 0 0 13 (0) 1 0.64 Re8-e7
00:00 0 0 14 (0) 1 0.84 Qb6-c5
00:00 0 0 17 (0) 1 0.88 Nd7-f6
00:00 0 0 18 (0) 1 1.01 Nd7-c5
00:00 0 0 19 (0) 1 1.04 Bh5-g6
00:00 0 0 33 (0) 1 1.08 Kg8-h7
00:00 0 0 36 (0) 1 1.13 Kg8-f8
00:00 0 0 102 (0) 2 0.93 Kg8-f8 Kh1-h2
00:00 0 0 379 (0) 3 1.20 Kg8-f8 Kh1-h2 Bh5-g6
00:00 0 0 1988 (391) 4 0.83 Kg8-f8 Kh1-g1 a5-a4 b3xa4
++ g8-h7
00:00 0 0 3025 (981) 4 0.87 Kg8-h7 Kh1-g1 Re8-g8 Kg1-f2
++ d7-c5
00:00 0 0 7015 (2310) 4 1.09 Nd7-c5 Re2-e1 Kg8-f8 a3-a4
++ a5-a4
00:00 0 0 9443 (2991) 4 1.13 a5-a4 g3-g4 Qb6-a6 Qc1-f1 Bh5-g6
00:00 0 0 16154 (3362) 5 0.92 a5-a4 Qc1-e1 Bh5xf3 Ne3-c4 Bf3xg2 Kh1xg2
++ g8-h7
00:00 0 0 26742 (8032) 5 1.03 Kg8-h7 Qc1-g1 Qb6-a6 Ne3-c4 b7-b5
++ g8-f8
00:00 0 0 31308 (8497) 5 1.07 Kg8-f8 Qc1-g1 Qb6-a6 Ne3-c4 b7-b5
++ d7-f6
00:01 0 0 42949 (11481) 5 1.13 Nd7-f6 Re2-e1 Kg8-f8 f3-f4 e5xf4 g3xf4
00:02 0 0 76128 (16961) 6 0.18 Nd7-f6 Qc1-e1 Bg5xe3 Re2xe3 Kg8-f8 Kh1-g1
++ g8-f8
00:02 0 0 98174 (24532) 6 0.62 Kg8-f8 Qc1-e1 a5-a4 Ne3-c4 Qb6-c5 Bb2-c3 a4xb3 c2
xb3
++ a5-a4
00:03 0 0 142640 (40777) 6 0.69 a5-a4 Qc1-e1 a4xb3 Ne3-f5 Kg8-f8 Qe1-b4 Qb6xb4 a
3xb4
00:06 0 0 281030 (55413) 7 0.92 a5-a4 Qc1-e1 a4xb3 Ne3-c4 Qb6-c7 c2xb3 Nd7-c5
00:10 0 0 544013 (91108) 8 0.70 a5-a4 Qc1-e1 a4xb3 Ne3-c4 Qb6-c7 c2xb3 Nd7-c5 Qe
1-b4
00:28 0 0 1753700 (112361) 9 0.89 a5-a4 Qc1-e1 a4xb3 Ne3-c4 Qb6-c7 c2xb3 Nd7-c5
Qe1-a5 Qc7-e7
00:48 0 0 3077596 (152700) 10 0.80 a5-a4 Qc1-e1 a4xb3 Ne3-c4 Qb6-c7 c2xb3 Nd7-c5
 Qe1-a5 Qc7xa5 Nc4xa5
04:12 0 0 17109023 (1591321) 11 0.68 a5-a4 b3-b4 f7-f5 e4xf5 e5-e4 f3-f4 Bg5xf4
Re2-e1 Bf4xg3 Ne3-c4 Bh5-d1
09:29 0 0 38492751 (3065091) 12 0.68 a5-a4 b3-b4 f7-f5 e4xf5 e5-e4 f3-f4 Bg5xf4
Re2-e1 Bf4xg3 Ne3-c4 Bh5-d1

So at 11 ply diep shows same mainline. Now this version of diep has not
much recapture extensions. I threw those out. Took too many nodes in
average middlegame to extend most recaptures.

---------------------------------------
--> re1e2 <--
---------------------------------------
 -34  T=21
Pa5a4 pb3b4 Pc6c5 qd2e2 Bg5e3n qe2e3B Pc5b4p bb2c1 Pb4a3p pg3g4 Bh5g6 ra1a3P
Qb6e3q
 3(4) -9  T=0
Pa5a4 pb3a4P Nd7c5
 4(5) -9  T=0
Pa5a4 pb3a4P Nd7c5
 5(5)[a4](-3) -3  T=0
Pa5a4 pb3a4P Nd7c5
 6(5)[a4](-26)[Qc5](-25) -25  T=0
Qb6c5 qc1g1 Qc5b5
 7(5) #[Qc5](-41)##########################[a4](-8)####### -8  T=3
Pa5a4 qc1e1 Pa4b3p ne3c4 Qb6c7 pc2b3P
 8(6)<ch> 's 60'
 #[a4](7)################################## 7  T=8
Pa5a4 pg3g4 Bh5g6 qc1e1 Pa4b3p ne3c4 Qb6c7 pc2b3P Nd7c5
 9(6) #[a4](15)################################## 15  T=39
Pa5a4 pg3g4 Bh5g6 pb3b4 Ph6h5 pg4h5P Bg6h5p qc1e1 Bg5f4 ne3f5
10(6) #[a4](-15)[a4](-15) -15v T=112
Pa5a4 pb3b4 Pf7f5 pe4f5P Pe5e4 pf3f4 Bg5f4p pg3f4B Bh5e2r qc1d2 Be2f3 ra1g1
10(6) #[a4](-20)[TIMEOUT][et3  946 sec]################################## -20
T=195
Pa5a4 pb3b4 Pf7f5 pe4f5P Pe5e4 pf3f4 Bh5e2r pf4g5B Nd7e5 pg5g6 Qb6c7 qc1b1 Kg8g7
pg3g4 Be2f3
---------------------------------------
-->  27. ..   a4 <-- 13/55:48
---------------------------------------

As you see Deep Blue gets it 1 ply sooner as DIEP. It gets it at 10 ply
versus DIEP 11 ply.

This amazing similarity in mainlines i already saw directly when i added
loads of what Hsu calls dangerous extensions.

Of course DIEP runs on a PC and it's 2001 now, so i don't want my search
to completely die because of extensions. So i limit them bigtime all.

So diep needs 1 ply extra usual as Deep Blue to get the same mainlines,
though usual it's the same.

This mainline is so important because it shows the tactical horizon effect
a4 b4 f5!? from black.

28..f5 is getting into the lost position as the rest is a forced win for
white. Computers do not know this of course. In the game i would be
sweatting like hell though with white. just 3 rows against the computer
has 5 rows!

Best Regards,
Vincent




This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.