Interesting position from Jouni Uski

[GeoffW]

Hi

[D]5K2/8/2qk4/2nPp3/3r4/6B1/B7/3R4 w - e6 0 1

Thought it was worth reposting this in programmers forum as this is a very interesting position for showing up engine bugs with FEN reading, ep captures and checks. This is a mate in 1 position due to the ep capture.

I was really surprised to find how many well known engines cant see the mate in 1

Ruffian, Crafty, Frenzee, Pharoan, Baron, Aice, Colossus, Homer2 to name but a few.

Happy bug hunting

Geoff

[NKOTB]

the old e.p. !! Delightful

[Alessandro Scotti]

No problem at all for Kiwi or Hamsters... this must be the first bug I do not have!

[Zlaire]

No problem for Mediocre either.

However are you sure it is not an interface problem? Meaning are you sure the actual position is passed to the engine? I know Arena has had problems with fen strings containing en passant squares (in the latest beta it can't input fen strings at all though ).

[GeoffW]

Hi Jonatan

Yes I am sure it is not an interface problem provided you use Arena version 1.1. Do not use the more recent betas as they have problems with fen handling.

For example as Alessandro pointed out, Hamsters and Kiwi work fine. I know, I tried them both with Arena, they escaped my hit list above

Geoff

[bob]

Crafty sees this just fine:

Code: Select all

White(1): move
              time surplus   0.00  time limit 22.50 (+0.00) (3:30)
              depth   time  score   variation (1)
                1     0.00  -2.24   1. Rxd4
                1     0.00   Mate   1. dxe6#
                1->   0.00   Mate   1. dxe6#
                2     0.00   Mate   1. dxe6#
                2->   0.00   Mate   1. dxe6#
                3     0.00   Mate   1. dxe6#
                3->   0.00   Mate   1. dxe6#
                4     0.00   Mate   1. dxe6#
                4->   0.00   Mate   1. dxe6#
              time=0.00  mat=-6  n=790  fh=100%  nps=1.0M
              ext-> check=18 1rep=0 mate=0 pp=0 reduct=121/0
              predicted=0  evals=424  50move=0  EGTBprobes=0  hits=0
              SMP->  splits=0  aborts=0  data=0/128  elap=0.00

Perhaps this is more of an operator error than a bug in multiple programs?

[bob]

you need to figure out why Crafty failed the test, when it works perfectly for any version I tried... There is definitely an interface or operator issue involved here.

[Alessandro Scotti]

you need to figure out why Crafty failed the test, when it works perfectly for any version I tried... There is definitely an interface or operator issue involved here.

At the time I did a quick test with Crafty on Mac OS X (it was version 19.13 IIRC, downloaded with Fink), just open a terminal, run crafty, setboard and analyze, and it failed to find the move.
What are the commands I have to give to obtain the panel you posted? After so much time with UCI/custom interface I forgot how to use XBoard protocol manually!

[bob]

you need to figure out why Crafty failed the test, when it works perfectly for any version I tried... There is definitely an interface or operator issue involved here.

At the time I did a quick test with Crafty on Mac OS X (it was version 19.13 IIRC, downloaded with Fink), just open a terminal, run crafty, setboard and analyze, and it failed to find the move.
What are the commands I have to give to obtain the panel you posted? After so much time with UCI/custom interface I forgot how to use XBoard protocol manually!

No idea what is wrong there. Here is 19.13 on my office box:

Code: Select all

crafty% crafty
unable to open book file [./book.bin].
book is disabled
unable to open book file [./books.bin].
EGTB access enabled
using tbpath=./TB
0 piece tablebase files found

Crafty v19.13 (1 cpus)

White(1): 5K2/8/2qk4/2nPp3/3r4/6B1/B7/3R4 w - e6 0 1
White(1): noise 0
noise level set to 0.
White(1): move
              clearing hash tables
              time surplus   0.00  time limit 30.00 (3:30)
              depth   time  score   variation (1)
                1     0.00  -1.16   1. Rxd4      
                1     0.00   Mate   1. dxe6#      
                1->   0.00   Mate   1. dxe6#      
                2     0.00   Mate   1. dxe6#      
                2->   0.00   Mate   1. dxe6#      
                3     0.00   Mate   1. dxe6#      
                3->   0.00   Mate   1. dxe6#      
                4     0.00   Mate   1. dxe6#      
                4->   0.00   Mate   1. dxe6#      
              time=0.00  cpu=100%  mat=-6  n=568  fh=100%  nps=10K
              ext-> chk=18 cap=0 pp=0 1rep=0 mate=0
              predicted=0  nodes=568  evals=3  50move=0
              endgame tablebase-> probes=0  hits=0
              SMP->  split=0  stop=0  data=0/64  cpu=0.00  elap=0.00

crafty% crafty
unable to open book file [./book.bin].
book is disabled
unable to open book file [./books.bin].
EGTB access enabled
using tbpath=./TB
0 piece tablebase files found

Crafty v19.13 (1 cpus)

White(1): 5K2/8/2qk4/2nPp3/3r4/6B1/B7/3R4 w - e6 0 1
White(1): noise 0
noise level set to 0.
White(1): analyze   
Analyze Mode: type "exit" to terminate.
              clearing hash tables
              time surplus   0.00  time limit 30.00 (3:30)
              depth   time  score   variation (1)
                1     0.00  -1.16   1. Rxd4      
                1     0.00   Mate   1. dxe6#      
                1->   0.00   Mate   1. dxe6#      
                2     0.00   Mate   1. dxe6#      
                2->   0.00   Mate   1. dxe6#      
                3     0.00   Mate   1. dxe6#      
                3->   0.00   Mate   1. dxe6#      
                4     0.00   Mate   1. dxe6#      
                4->   0.00   Mate   1. dxe6#      
              time=0.00  cpu=100%  mat=-6  n=568  fh=100%  nps=10K
              ext-> chk=18 cap=0 pp=0 1rep=0 mate=0
              predicted=0  nodes=568  evals=3  50move=0
              endgame tablebase-> probes=0  hits=0
              SMP->  split=0  stop=0  data=0/64  cpu=0.01  elap=0.00

So it doesn't appear to fail here no matter how you try it. You do have to use the "noise 0" command or it will just annouce the mate without showing anything since the PV takes just 500 nodes total to search...

[Alessandro Scotti]

I have tried again with the same commands posted above and obtained the same result you posted (i.e. correct mate found)... who knows what I did when I tried that for the first time!