Re: Gothmog - the best tactician - PETv1

[Gábor Szõts]

Geschrieben von:/Posted by: Gábor Szõts at 05 April 2004 08:03:04:

Hi, above all Peter Fendrich,
You asked whether PETv1 test results of various engines are available. I have done this test using several engines. To post results here would be a bit difficult, but I'm going to send them to anyone who asks.
BTW, requirement is 1 minute per position.
Gábor

[Manfred Meiler]

Geschrieben von:/Posted by: Manfred Meiler at 05 April 2004 09:23:17:
Als Antwort auf:/In reply to: Re: Gothmog - the best tactician - PETv1 geschrieben von:/posted by: Gábor Szõts at 05 April 2004 08:03:04:

Hi, above all Peter Fendrich,
You asked whether PETv1 test results of various engines are available. I have done this test using several engines. To post results here would be a bit difficult, but I'm going to send them to anyone who asks.
BTW, requirement is 1 minute per position.
Gábor

Hi Gábor,
you're right, Gothmog is really a great tactician.
In the 38 test positions "king attack" of the test suite "Weltmeister-Test" (WM-Test) http://www.computerschach.de/test/index.htm the former Gothmog version 0.4.5 scored (with 27 solutions) at the same level with e.g. Junior 8, Ruffian 2.0.0 or Shredder 7.0 (http://f23.parsimony.net/forum50826/messages/92389.htm).
Manfred

[Tord Romstad]

Geschrieben von:/Posted by: Tord Romstad at 05 April 2004 11:01:02:
Als Antwort auf:/In reply to: Re: Gothmog - the best tactician - PETv1 geschrieben von:/posted by: Manfred Meiler at 05 April 2004 09:23:17:

you're right, Gothmog is really a great tactician.

This is, to a great extent, just an illusion. As I explained in the CCC a
few days ago, Gothmog is "cheating". It has huge speculative king safety
terms which often makes it guess that some sacrifical attack works a couple
of plies before it searches deeply enough to calculate the whole combination.
Considering that it is such a slow engine, Gothmog is a somewhat better
tactician than you would expect. The last couple of versions are probably
tactically stronger than the majority of amateur engines. But even today,
Gothmog is far behind the best engines (amateurs as well as professionals)
at tactics, and calling it a "great tactician" is a considerable exaggeration.
Tord

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 05 April 2004 11:08:08:
Als Antwort auf:/In reply to: Re: Gothmog - the best tactician - PETv1 geschrieben von:/posted by: Manfred Meiler at 05 April 2004 09:23:17:

Hi, above all Peter Fendrich,
You asked whether PETv1 test results of various engines are available. I have done this test using several engines. To post results here would be a bit difficult, but I'm going to send them to anyone who asks.
BTW, requirement is 1 minute per position.
Gábor

Hi Gábor,
you're right, Gothmog is really a great tactician.
In the 38 test positions "king attack" of the test suite "Weltmeister-Test" (WM-Test) http://www.computerschach.de/test/index.htm the former Gothmog version 0.4.5 scored (with 27 solutions) at the same level with e.g. Junior 8, Ruffian 2.0.0 or Shredder 7.0 (http://f23.parsimony.net/forum50826/messages/92389.htm).
Manfred

King attack is not tactics.
Moves can be found for positional reasons and if you have only
find sacrifices that work and not avoid sacrifices that do not work then the test is misleading even in testing positional understanding.
tactics is about seeing something in the tree faster than the oppoent.
I think that correct test for tactics is to ask programs not for the right move but how much time they need to consider a set of specific lines.
Of course programmers need to modify their program to check at every node if they consider one of the specific lines.
Uri

[Peter Fendrich]

Geschrieben von:/Posted by: Peter Fendrich at 05 April 2004 14:42:03:
Als Antwort auf:/In reply to: Re: Gothmog - the best tactician - PETv1 geschrieben von:/posted by: Gábor Szõts at 05 April 2004 08:03:04:

Hi, above all Peter Fendrich,
You asked whether PETv1 test results of various engines are available. I have done this test using several engines. To post results here would be a bit difficult, but I'm going to send them to anyone who asks.
BTW, requirement is 1 minute per position.
Gábor

Yes I'm interested!
Why not post them on your homepage?
/Peter