Poliglot and List 512 (part 2)

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 31 March 2004 12:07:18:

List 512 still has problems under Poliglot (dated 31st March)
From winboard.debug
====================
548930

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 12:20:07:
Als Antwort auf:/In reply to: Poliglot and List 512 (part 2) geschrieben von:/posted by: Igor Gorelikov at 31 March 2004 12:07:18:

List 512 still has problems under Poliglot (dated 31st March)
From winboard.debug
====================
548930 548930 Fatal Error: Error: second chess program (polyglot.exe List_512.ini) exited unexpectedly
From Poliglot's log (for List 512)
==================================
< ENGINE bestmove b6b8 ponder
engine_step(): not a move "b6b8 ponder"

Thanks but to be 100% sure of what the problem is, I would like to check with *List* log file if possible.
If lists "admits" sending "move (xxx) ponder" with nothing after ponder, then IMO it is definately a List bug.
But first we need to make sure it is.
Fabien.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 12:25:21:
Als Antwort auf:/In reply to: Re: Polyglot and List 512 (part 2) geschrieben von:/posted by: Fabien Letouzey at 31 March 2004 12:20:07:

If lists "admits" sending "move (xxx) ponder" with nothing after ponder, then IMO it is definately a List bug.

I meant "bestmove (xxx) ponder".
Fabien.

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 31 March 2004 12:26:48:
Als Antwort auf:/In reply to: Re: Polyglot and List 512 (part 2) geschrieben von:/posted by: Fabien Letouzey at 31 March 2004 12:20:07:

Thanks but to be 100% sure of what the problem is, I would like to check with *List* log file if possible.

List has no log file.
Igor

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 12:37:02:
Als Antwort auf:/In reply to: Re: Polyglot and List 512 (part 2) geschrieben von:/posted by: Igor Gorelikov at 31 March 2004 12:26:48:

Thanks but to be 100% sure of what the problem is, I would like to check with *List* log file if possible.

List has no log file.
Igor

I see.
You can understand my problem. I suspect a bug in List but I can't trust PolyGlot log file 100% because there were I/O problems with Windows GCC that I am not sure I have fully solved.
If using List with WB is important for you, then you must go further:
1) use List in a UCI interface that has log files
2) play games/matches/whatever
3) locate a line with "bestmove (xxx) ponder" (nothing after ponder) in the log file
4) if you find one, then you should contact List's author.
In any case, let me know.
I *might* add an *optional* work around for this problem but this is against my usual programming policy (bugs must be fixed in place, not in other programs).
If List is important to you, I will add a List-specific option, but only after you follow steps 1) to 4) above.
Sorry I can't be more helpful (although fixing bugs in engines is the best way in the long run),
Fabien.

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 31 March 2004 13:29:09:
Als Antwort auf:/In reply to: Re: Polyglot and List 512 (part 2) geschrieben von:/posted by: Fabien Letouzey at 31 March 2004 12:37:02:

Now Arena 1.02 is used.
From Arena.debug:
=================
734851(1:info score cp 32765 depth 31 nodes 2862 time 50 pv d8c8
734872(1:bestmove d8c8 ponder
734872*1*Found move:Qd8xc8
735110********Game end: 1-0 {Mate}
735110********Last game won by List 5.12 with White in 16 moves
735110********List 5.12 - S-chess 1.4: 1-0-0
Igor

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 13:44:38:
Als Antwort auf:/In reply to: Re: Polyglot and List 512 (part 2) geschrieben von:/posted by: Igor Gorelikov at 31 March 2004 13:29:09:

Now Arena 1.02 is used.
From Arena.debug:
=================
734851(1:info score cp 32765 depth 31 nodes 2862 time 50 pv d8c8
734872(1:bestmove d8c8 ponder
734872*1*Found move:Qd8xc8
735110********Game end: 1-0 {Mate}
735110********Last game won by List 5.12 with White in 16 moves
735110********List 5.12 - S-chess 1.4: 1-0-0
Igor

That's a proof, good job!
Please report this to List's author, although after the problems he had I doubt he will care ...
If you consider it important, I will add an optional work around.
Fabien.

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 31 March 2004 14:08:02:
Als Antwort auf:/In reply to: Re: Polyglot and List 512 (part 2) geschrieben von:/posted by: Fabien Letouzey at 31 March 2004 13:44:38:

If you consider it important, I will add an optional work around.
Fabien.

If you can add adjudication to Poliglot (with a settable losing score) it will be useful not only for List but for many TDs.
Also it will be solution for List 512.
Igor

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 14:19:33:
Als Antwort auf:/In reply to: Re: Polyglot and List 512 (part 2) geschrieben von:/posted by: Igor Gorelikov at 31 March 2004 14:08:02:

If you consider it important, I will add an optional work around.
Fabien.

If you can add adjudication to Poliglot (with a settable losing score) it will be useful not only for List but for many TDs.
Also it will be solution for List 512.
Igor

You are putting me into a difficult position.
On one hand the output from list is not "legal" UCI.
On the other hand it's obvious what its meaning is, List wants to play some move xxx and does not send anything to ponder on which in itself is OK (although it should be formulated differently).
So far PolyGlot is not ready at all for tournament play, and should of course not be used for serious matches as yet.
If you want I can add a work around for List, but you will have to activate it in list.ini
Also, it is important for me that PolyGlot stays "impartial" regarding the game result. So it should not adjudicate in any way. As of now, it only reports draws by material/50-move/3-rep (plus mate/stalemate of course).
Or maybe you can explain your request in a different way?
Fabien.

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 31 March 2004 14:50:05:
Als Antwort auf:/In reply to: Re: Polyglot and List 512 (part 2) geschrieben von:/posted by: Fabien Letouzey at 31 March 2004 14:19:33:

If you want I can add a work around for List, but you will have to activate it in list.ini
Also, it is important for me that PolyGlot stays "impartial" regarding the game result. So it should not adjudicate in any way.

It would be good.
Adjudication is very important (especially for long time control). It saves many houres of engine testing.
And you may do it switchable to let the user decide.
Igor

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 14:52:56:
Als Antwort auf:/In reply to: Adjudication geschrieben von:/posted by: Igor Gorelikov at 31 March 2004 14:50:05:

If you want I can add a work around for List, but you will have to activate it in list.ini
Also, it is important for me that PolyGlot stays "impartial" regarding the game result. So it should not adjudicate in any way.

It would be good.
Adjudication is very important (especially for long time control). It saves many houres of engine testing.
And you may do it switchable to let the user decide.
Igor

Please define "Adjudication" clearly; when should it trigger?
Fabien.

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 31 March 2004 15:10:50:
Als Antwort auf:/In reply to: Re: Adjudication geschrieben von:/posted by: Fabien Letouzey at 31 March 2004 14:52:56:

If you want I can add a work around for List, but you will have to activate it in list.ini
Also, it is important for me that PolyGlot stays "impartial" regarding the game result. So it should not adjudicate in any way.

It would be good.
Adjudication is very important (especially for long time control). It saves many houres of engine testing.
And you may do it switchable to let the user decide.
Igor

Please define "Adjudication" clearly; when should it trigger?
Fabien.

Adjudication is very similar to the resign option which many engines use.
For instance, Crafty resign after 5 consecutive moves with score -x (I set it to -6).
Adjudication can have two paramenters: negative score (usually settable) and the number of consecutive moves (which is optional and can be hardcoded to 3-5).
Igor

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 15:16:14:
Als Antwort auf:/In reply to: Re: Adjudication geschrieben von:/posted by: Igor Gorelikov at 31 March 2004 15:10:50:

If you want I can add a work around for List, but you will have to activate it in list.ini
Also, it is important for me that PolyGlot stays "impartial" regarding the game result. So it should not adjudicate in any way.

It would be good.
Adjudication is very important (especially for long time control). It saves many houres of engine testing.
And you may do it switchable to let the user decide.
Igor

Please define "Adjudication" clearly; when should it trigger?
Fabien.

Adjudication is very similar to the resign option which many engines use.
For instance, Crafty resign after 5 consecutive moves with score -x (I set it to -6).
Adjudication can have two paramenters: negative score (usually settable) and the number of consecutive moves (which is optional and can be hardcoded to 3-5).
Igor

This sort of adjudication, I see. Not urgent I am afraid. This week we are testing PolyGlot compatibility with Windows, xboard GUIs (not just WinBoard) and UCI engines (not just UCI-only engines).
Feature requests will be considered later, when major problems stop showing up.
Also, adjudication is a lot of work; I have no plan for it in the near future I am afraid ...
Fabien.

[Igor Gorelikov]

Geschrieben von:/Posted by: Igor Gorelikov at 31 March 2004 15:40:28:
Als Antwort auf:/In reply to: Re: Adjudication geschrieben von:/posted by: Fabien Letouzey at 31 March 2004 15:16:14:

Feature requests will be considered later, when major problems stop showing up.
Also, adjudication is a lot of work; I have no plan for it in the near future I am afraid ...
Fabien.

Off course, it is not the first thing to do!
Igor