Avoid move test

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

Moderators: hgm, Rebel, chrisw

gladius
Posts: 568
Joined: Tue Dec 12, 2006 10:10 am
Full name: Gary Linscott

Avoid move test

Post by gladius »

Based on the rook odds game with GM Kaufmann, one of the playouts of Houdini vs. GarboChess JS ended up at the following position:

[d]r4r2/p1p1k2p/1p4p1/2qnBBR1/3p4/7P/P3QP1K/8 b - -[/d]

GarboChess played d3 here, leading to being mated in 10, but it takes a long time to figure out that it is mated :). Even strong programs consider d3 to begin with it appears though, so I don't feel too bad.
Top
jdart
Posts: 4367
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: Avoid move test

Post by jdart »

I don't think this is very hard.

Arasan, recent build, switches off d3 after 0.19 seconds (depth 7):

Using book file book.bin
am d3
0 0.07 Rxf5! +6.48 39837
0 0.07 Rxf5 +10.02 39841
1 0.07 -- +9.52 39919
1 0.07 Rxf5 +2.28 39921
1 0.07 d3 +5.98 39922
2 0.07 d3 +5.98 39961
3 0.07 d3 +5.98 40001
4 0.07 d3! +6.48 43499
4 0.08 d3 +6.64 48510
5 0.10 d3 +6.64 59923
6 0.13 d3 +6.46 83748
7 0.18 -- +5.96 149640
7 0.19 d3 -Mate10 152716
7 0.57 Ne3 +1.76 659187
7 0.80 Qb4 +2.36 992321
7 0.86 Kd8 +5.30 1057473
8 1.10 Kd8 +5.04 1277243
9 1.22 Kd8 +4.84 1409997
10 1.50 Kd8 +4.76 1735171
Top
User avatar
beachknight
Posts: 3533
Joined: Tue Jan 09, 2007 8:33 pm
Location: Antalya, Turkey

Re: Avoid move test

Post by beachknight »

A new version of Garbo in sight, Gary?

:)
hi, merhaba, hallo HT
Top
gladius
Posts: 568
Joined: Tue Dec 12, 2006 10:10 am
Full name: Gary Linscott

Re: Avoid move test

Post by gladius »

It's not super hard, but it's impressive that Arasan spots the mate at depth 7!

There is a long sequence of checks+material sacrifices in the mate, most programs don't see it right away.
Top
gladius
Posts: 568
Joined: Tue Dec 12, 2006 10:10 am
Full name: Gary Linscott

Re: Avoid move test

Post by gladius »

beachknight wrote:A new version of Garbo in sight, Gary?
Only the Javascript version. The native UCI engine space is pretty crowded these days, I'm not dedicated enough to compete there :).
Top
PawnStormZ
Posts: 880
Joined: Mon Feb 15, 2010 6:43 am

Avoid move test

Post by PawnStormZ »

 
   Houdini 1.03a sees the mate instantly!!!

[Mate in 10] d=29 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.11) 126995kN

[Mate in 10] d=28 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.07) 81784kN

[Mate in 10] d=27 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.05) 51532kN

[Mate in 10] d=26 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.03) 33415kN

[Mate in 10] d=25 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.02) 20859kN

[Mate in 10] d=24 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.01) 12483kN

[Mate in 10] d=23 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.01) 7677kN

[Mate in 10] d=22 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 4372kN

[Mate in 10] d=21 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 2478kN

[Mate in 10] d=20 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 1416kN

[Mate in 10] d=19 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 762kN

[Mate in 10] d=18 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 420kN

[Mate in 10] d=17 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 218kN

[Mate in 10] d=16 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 108kN

[Mate in 10] d=15 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 68kN

[Mate in 10] d=14 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 50kN

[Mate in 10] d=13 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 41kN

[Mate in 10] d=12 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 38kN

[Mate in 10] d=11 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.h4 Kh5 10.Qg6 Kxh4 11.Qg4# (0:00.00) 38kN

[+13.85] d=9 2.Bf6 Kxf6 3.Rxg6 hxg6 4.Qe6 Kg5 5.Qxg6 Kf4 6.Qg4 Ke5 7.Qe4 Kf6 8.Qe6 Kg5 9.Qg6 Kf4 (0:00.00) 36kN
 
Top
Cubeman
Posts: 644
Joined: Fri Feb 02, 2007 3:11 am
Location: New Zealand

Re: Avoid move test

Post by Cubeman »

I set position up before the move d3 was played and many engines want to play 1..d3?? The worst that I've seen was the new iphone Glaurung 2.0 which took to D=13 and over 2 minutes to finally regect 1..d3.Meanwhile one of the fastest was Chess Genius which spots it at low depths, Shredder was fast too.Naum 4.2 and Hiarcs 13.1 were o.k.But Crafty 23.0 was even faster.
Top
User avatar
michiguel
Posts: 6401
Joined: Thu Mar 09, 2006 8:30 pm
Location: Chicago, Illinois, USA

Re: Avoid move test

Post by michiguel »

gladius wrote:Based on the rook odds game with GM Kaufmann, one of the playouts of Houdini vs. GarboChess JS ended up at the following position:

[d]r4r2/p1p1k2p/1p4p1/2qnBBR1/3p4/7P/P3QP1K/8 b - -[/d]

GarboChess played d3 here, leading to being mated in 10, but it takes a long time to figure out that it is mated :). Even strong programs consider d3 to begin with it appears though, so I don't feel too bad.

It takes 6.3 seconds to Gaviota to see problems with d3 (with a massive fail low) (AMD 2.4 Ghz, one core)

Code: Select all

setboard r4r2/p1p1k2p/1p4p1/2qnBBR1/3p4/7P/P3QP1K/8 b - -
d
+-----------------+
| r . . . . r . . | [Black]
| p . p . k . . p |
| . p . . . . p . |
| . . q n B B R . |    Castling: 
| . . . p . . . . |    ep: -
| . . . . . . . P |
| P . . . Q P . K |
| . . . . . . . . |
+-----------------+

analyze
********* Starts iterative deepening, thread = 0
set timer to infinite
       306   1:      0.0    +4.31  d3
       960   2       0.0      :-(  
      1935   2:      0.0    +3.85  d3 2.Bd6+ Kd8
      3740   3       0.0      :-)  d3
      5705   3       0.0      :-)  d3
      9809   3:      0.1    +5.55  d3 2.Bd6+ Kd8 3.Bxc5 dxe2
     19813   4       0.1      :-)  d3
     34743   4:      0.2    +6.69  d3 2.Qxd3 Rxf5 3.Rxf5 gxf5 4.Qxf5
     73965   5:      0.4    +6.69  d3 2.Qxd3 Rxf5 3.Rxf5 gxf5 4.Qxf5
    211656   6       1.0    +6.49  d3 2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8 6.Qxh7 Qxf2+ 7.Kh1
    238187   6:      1.1    +6.49  d3 2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8 6.Qxh7 Qxf2+ 7.Kh1
    511994   7       2.1    +6.58  d3 2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8
    684642   7:      2.8    +6.58  d3 2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8
   1594208   8       6.3      :-(  d3
   1934410   8       7.6      :-(  
   2026469   8       7.9      :-(  d3
   2585667   8      10.1      :-(  
   2586302   8      10.1      :-(  d3
  14629940   8      56.4    +3.88  Kd8 2.Be4 Nc3 3.Qd3 Nxe4 4.Qxe4 Kd7
                                   5.Qg4+ Rf5 6.Bxd4 Qd6+ 7.Kh1
  15029189   8:     58.0    +3.88  Kd8 2.Be4 Nc3 3.Qd3 Nxe4 4.Qxe4 Kd7
                                   5.Qg4+ Rf5 6.Bxd4 Qd6+ 7.Kh1
ex  17829912   9      68.8    +3.92  Kd8 2.Be4 Nc3 3.Qd3 Nxe4 4.Qxe4 Kd7
                                   5.Qg4+ Rf5 6.Bxd4 Qd6+ 7.Kg1
If d3 is played, it sees the mate in 4 seconds

Miguel

Code: Select all

d3
analyze
********* Starts iterative deepening, thread = 0
set timer to infinite
       493   1:      0.0    -3.85  2.Bd6+ Kd8
       998   2       0.0      :-)  2.Bf6+
      1136   2       0.0      :-(  
      1714   2:      0.0    -5.55  2.Bd6+ Kd8 3.Bxc5 dxe2
      1801   3       0.0      :-)  2.Bf6+
      2676   3       0.0      :-(  
     13190   3:      0.1    -6.69  2.Qxd3 Rxf5 3.Rxf5 gxf5 4.Qxf5
     17994   4       0.1      :-)  2.Bf6+
     39072   4:      0.2    -6.69  2.Qxd3 Rxf5 3.Rxf5 gxf5 4.Qxf5
     41249   5       0.2      :-)  2.Bf6+
     87431   5:      0.4    -6.49  2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8 6.Qxh7 Qxf2+ 7.Kh1
     96698   6       0.5    -6.58  2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8
     96704   6       0.5      :-)  2.Bf6+
    112815   6       0.5      :-(  2.Bf6+
    113109   6       0.5    -6.58  2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8
    117625   6:      0.6    -6.58  2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8
    138090   7       0.7    -6.58  2.Qe4 gxf5 3.Rxf5 Kd7 4.Qxd3 Rxf5
                                   5.Qxf5+ Kd8
    138138   7       0.7      :-)  2.Bf6+
    138295   7       0.7      :-)  2.Bf6+
    152593   7       0.7    +0.00  2.Bf6+ Kxf6 3.Rxg6+ hxg6 4.Qe6+ Kg5
                                   5.Qxg6+ Kf4 6.Qg4+ Ke5 7.Qe4+ Kf6
                                   8.Qe6+ Kg5 9.Qg6+
    269825   7:      1.2    +0.00  2.Bf6+ Kxf6 3.Rxg6+ hxg6 4.Qe6+ Kg5
                                   5.Qxg6+ Kf4 6.Qg4+ Ke5 7.Qe4+ Kf6
                                   8.Qe6+ Kg5 9.Qg6+
    269944   8       1.2      :-)  2.Bf6+
   1025976   8       4.0  +Mat_10  2.Bf6+ Kxf6 3.Rxg6+ hxg6 4.Qe6+ Kg5
                                   5.Qxg6+ Kf4 6.Qg4+ Ke5 7.Qe4+ Kf6
                                   8.Qe6+ Kg5 9.h4+ Kh5 10.Qg6+ Kxh4
                                   11.Qh6#
  10009580   8:     38.2  +Mat_10  2.Bf6+ Kxf6 3.Rxg6+ hxg6 4.Qe6+ Kg5
                                   5.Qxg6+ Kf4 6.Qg4+ Ke5 7.Qe4+ Kf6
                                   8.Qe6+ Kg5 9.h4+ Kh5 10.Qg6+ Kxh4
                                   11.Qh6#
Top
User avatar
silentshark
Posts: 327
Joined: Sat Mar 27, 2010 7:15 pm

Re: Avoid move test

Post by silentshark »

Well, my prog really struggles to avoid ..d3 for a minute or two. I will see if any of the usual suspects (lmr or nullmove) are causing this..

I don't do checks in qsearch, maybe that would help?
Top
User avatar
silentshark
Posts: 327
Joined: Sat Mar 27, 2010 7:15 pm

Re: Avoid move test

Post by silentshark »

silentshark wrote:Well, my prog really struggles to avoid ..d3 for a minute or two. I will see if any of the usual suspects (lmr or nullmove) are causing this..

I don't do checks in qsearch, maybe that would help?
:oops:

issue with lmr - I was reducing for checking moves.. now my prog sees the issue with ..d3 pretty much instantly, and switches to another move after a couple of seconds.
Top

Return to “Computer Chess Club: General Topics”