Wb2Uci

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 23 September 2002 12:31:45:

Hi.
I'm doing the final tests for the next release of Wb2Uci adapter and need some input for which engine to test.
Currently my test is for:
Crafty – Because it gives the user easier access to all parameters from inside Arena/Fritz/Shredder.
Yace – Because I can compare uci and wb.
Gandalf 5.x – Because other do the test.
The King 3.xx – Because I get too much email if this don’t work.
PolarEngine – Because this is the reason for the adapter.
Why did I make an adapter instead of implementing the uci-protocol inside my engine?
To give the average user access to the wonderful world of Winboard-engines through the GUI they are used to play with.
As you see of my list there is very few engines that fit my original intention so I would like it if there is some engines out there that is suited for the average user.
Who is the average user?
Well a strong home player would be in the range of 1200-1300 elo points so anything from this and downwards is interesting.
The average club-player is at 1400 but here you would find many players in the range from 1000-2000.
My elo number is calculated as an extention to Fide's elo number. I checked the active Scandinavian player who figured in both Fide's list and the national list and found that the Danish and Norwegian list is quite accurate. The Swedish elo was ca. 50 above this extended list. For other country you have to calculate this yourself.
What about these elo-list for engines? I can only speak for myself, but I have to look for equal opponents in the range 4-500 elo points above my own rating if I want to play a standard game. Not all list are equal but if PolarChess is above 1800 you have found such a list.
All engines that have a possibility to alter search/eval parameters are interesting, especially if they have a method to change this on the fly.
This doesn't need to be that all commands are accessible from the console interface. A method that is just as good is to implement a 'readini' command so there could be several ini-files for different style/strength.
That I see in the future is a lot of (preset) personalities for different engines so that the user could select Alfred Crafty, Sven Terra or maybe Polaren Per to play against with just a mouse click.
Another command that I think is necessary for human-comp play is the 'move now' command. To not let the user quit, start a new game, and etc… anytime he want would not bring him back to this engine. The 'move now' command doesn't need to be implemented correct for this to work. If it answers a command under search with a move, so does not an additional 'illegal move' harm this.
Does the adapter reduce the strength of the engine?
Well, there is some processing in it so a little cpu power will be taken. Maybe 1-3% but this is very little if you should try to make an elo estimate out of it (1 point?). The gui does a lot more work so I don’t see this as a problem.
What about book learning? It will get some result back if it loses but many results isn't possible without starting to guess. Then another question would be: how many games do you think a player would play against the engine, this is not an engine-engine match with several 100 games. A game takes maybe 2-4 hour with full concentration.
Positional learning should maybe be more important for these wanted games because if the engine were used for learning he would most probably want to repeat openings. The nice thing here would be if the engine could change the play a little after the opening. An example could be Hyatt's method from ICCA 22/4 where he collect scores after the search and put it in a permanent hash table for use on the next games. I have not come to try this myself but the idea seems to let the engine find better moves the next time it plays the same opening, just as a human.
Odd Gunnar

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 23 September 2002 13:02:56:
Als Antwort auf: / In reply to: Wb2Uci geschrieben von: / posted by: Odd Gunnar Malin at 23 September 2002 12:31:45:


Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 23 September 2002 14:00:38:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Matthias Gemuh at 23 September 2002 13:02:56:

Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

Is there any problem with the uci version?
Anyway I did a run with timecontrol 1'/25 + 1'/15 +1'/g just now and the adapterversion won 4-0 against the uci version.
My eng file was:
[ENGINE]
Name=YaceWB 0.99.56
Author=Dieter Buerssner
Filename=Wb2Uci.exe
[OPTIONS]
Program=yacewb.exe
HashCommand=hash %iM
Hash=1
Logfile=false
Visible=Ponder,Hash,Logfile,Extras,OwnBook
[EXTRAS]
Clear Hash=reset|button
NalimovPath=tbldir|string|.\
NalimovCache=egtb_cache %iM|spin|1|1|16
Book=book|string|dbbook.bin
Yace log=auto_log %s|check|false|0|1
Inside Arena I set ponder off and hashsize=12MB, else as default for each version.
The adapter version had a little more time on average for the last moves before timecontrol.
The result is of course not much to see to but at least the adapter doesn't weaken the engine much.
You have to use the 1.3beta version to get Yace to play without stop.
Odd Gunnar

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 23 September 2002 14:25:15:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Odd Gunnar Malin at 23 September 2002 14:00:38:

Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

Is there any problem with the uci version?
Anyway I did a run with timecontrol 1'/25 + 1'/15 +1'/g just now and the adapterversion won 4-0 against the uci version.

The last time I tested it under Fritz7 GUI at 10'/40+5'/20+6',
engines (Holmes+Cerebro) would play normally during first 40 moves and after move 60, but use only about 0 seconds in 2nd time interval (moves 41 ... 60).
I will test this behaviour again.
/Matthias.

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 23 September 2002 14:45:20:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Matthias Gemuh at 23 September 2002 14:25:15:

Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

Is there any problem with the uci version?
Anyway I did a run with timecontrol 1'/25 + 1'/15 +1'/g just now and the adapterversion won 4-0 against the uci version.

The last time I tested it under Fritz7 GUI at 10'/40+5'/20+6',
engines (Holmes+Cerebro) would play normally during first 40 moves and after move 60, but use only about 0 seconds in 2nd time interval (moves 41 ... 60).
I will test this behaviour again.
/Matthias.

This could well be, I have to test them separate to see if they need a special setup.
I have tree setup for the level command.
;Default
LevelExtend=progressive
;If this doesn't helt try:
LevelExtend=strict
;If this neither help try:
LevelExtend=failsafe
If nothing help there could be a bug in the adapter as well
I will check this up later.
Odd Gunnar

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 23 September 2002 17:19:51:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Matthias Gemuh at 23 September 2002 14:25:15:

Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

Is there any problem with the uci version?
Anyway I did a run with timecontrol 1'/25 + 1'/15 +1'/g just now and the adapterversion won 4-0 against the uci version.

The last time I tested it under Fritz7 GUI at 10'/40+5'/20+6',
engines (Holmes+Cerebro) would play normally during first 40 moves and after move 60, but use only about 0 seconds in 2nd time interval (moves 41 ... 60).
I will test this behaviour again.
/Matthias.

For Holmes you need LevelExtend=strict
Ex.
[ENGINE]
Name=Holmes 0.80
Author=Andreas Herrmann
Filename=Wb2Uci.exe
[OPTIONS]
Program=Holmes080.exe xboard
OwnBook=true
Edit=edit
LevelExtend=strict
Visible=Logfile,OwnBook,Ponder

As you see I didn't have the latest version here, but my bet is that this isn't changed.
That you could do is to check up the readmefile and 'help' from command prompt to see if there could be more parameters to set. f.ex. hash. I think I saw that it had ponder in latest version so I added this to the visible line.
I don't know about the other engine, is it private/commercial?
You could try the same parameter (LevelExtend=strict) to see.
About half the engines need this.
Odd Gunnar

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 23 September 2002 19:38:06:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Odd Gunnar Malin at 23 September 2002 17:19:51:

Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

Is there any problem with the uci version?
Anyway I did a run with timecontrol 1'/25 + 1'/15 +1'/g just now and the adapterversion won 4-0 against the uci version.

The last time I tested it under Fritz7 GUI at 10'/40+5'/20+6',
engines (Holmes+Cerebro) would play normally during first 40 moves and after move 60, but use only about 0 seconds in 2nd time interval (moves 41 ... 60).
I will test this behaviour again.
/Matthias.

For Holmes you need LevelExtend=strict
Ex.
[ENGINE]
Name=Holmes 0.80
Author=Andreas Herrmann
Filename=Wb2Uci.exe
[OPTIONS]
Program=Holmes080.exe xboard
OwnBook=true
Edit=edit
LevelExtend=strict

Maybe I was abit fast here, when I try with exactly your timecontrol I have to use LevelExtend=failsafe Eg. on t/40+t/20+t/g
I will study this a bit more.
Odd Gunnar

[Andreas Herrmann]

Geschrieben von: / Posted by: Andreas Herrmann at 23 September 2002 20:13:14:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Odd Gunnar Malin at 23 September 2002 19:38:06:

Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

Is there any problem with the uci version?
Anyway I did a run with timecontrol 1'/25 + 1'/15 +1'/g just now and the adapterversion won 4-0 against the uci version.

The last time I tested it under Fritz7 GUI at 10'/40+5'/20+6',
engines (Holmes+Cerebro) would play normally during first 40 moves and after move 60, but use only about 0 seconds in 2nd time interval (moves 41 ... 60).
I will test this behaviour again.
/Matthias.

For Holmes you need LevelExtend=strict
Ex.
[ENGINE]
Name=Holmes 0.80
Author=Andreas Herrmann
Filename=Wb2Uci.exe
[OPTIONS]
Program=Holmes080.exe xboard
OwnBook=true
Edit=edit
LevelExtend=strict

Maybe I was abit fast here, when I try with exactly your timecontrol I have to use LevelExtend=failsafe Eg. on t/40+t/20+t/g
I will study this a bit more.
Odd Gunnar

Hi Odd Gunnar,
is it possible to use wb2uci under Shredder 5? That's the only professional GUI that seems using the UCI protocoll, i have. Because i want to check your wb2uci by myself.
Andreas

[Thomas Lagershausen]

Geschrieben von: / Posted by: Thomas Lagershausen at 23 September 2002 21:04:06:
Als Antwort auf: / In reply to: Wb2Uci geschrieben von: / posted by: Odd Gunnar Malin at 23 September 2002 12:31:45:

First thanks a lot for this amazing tool.
It makes that i can analyze with this fantastic king-engine under Fritz and the shredder-gui.
Now for my problem.
With all versions of your adapter the king-engine stops always(!) thinking in analyze-mode after between about 44 and 72 minutes(hardware:Athlon 1,5 Ghz).
I am a tournament player and love it to let two programs run the whole night on a position.
It would be a dream if you can find the problem.
With the inbetween ini it works, but the WB2Uci tool is off course better.
Greetings
Thomas Lagershausen

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 23 September 2002 22:29:47:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Andreas Herrmann at 23 September 2002 20:13:14:

Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

Is there any problem with the uci version?
Anyway I did a run with timecontrol 1'/25 + 1'/15 +1'/g just now and the adapterversion won 4-0 against the uci version.

The last time I tested it under Fritz7 GUI at 10'/40+5'/20+6',
engines (Holmes+Cerebro) would play normally during first 40 moves and after move 60, but use only about 0 seconds in 2nd time interval (moves 41 ... 60).
I will test this behaviour again.
/Matthias.

For Holmes you need LevelExtend=strict
Ex.
[ENGINE]
Name=Holmes 0.80
Author=Andreas Herrmann
Filename=Wb2Uci.exe
[OPTIONS]
Program=Holmes080.exe xboard
OwnBook=true
Edit=edit
LevelExtend=strict

Maybe I was abit fast here, when I try with exactly your timecontrol I have to use LevelExtend=failsafe Eg. on t/40+t/20+t/g
I will study this a bit more.
Odd Gunnar

Hi Odd Gunnar,
is it possible to use wb2uci under Shredder 5? That's the only professional GUI that seems using the UCI protocoll, i have. Because i want to check your wb2uci by myself.
Andreas

Yes I would think so. I have been using it under Shredder 6 (Classic) and the april version (1.2) is tested for it. I have not come so long in my testing on the new version that I have tested it in Shredder.
I guess I expect it to run smoothly in there so I leave it to the final test 1 minute before I release it (remove the beta).
You could also try it on Arena. There you can match it against itself as a wb-engine to see any differences.
Odd Gunnar

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 23 September 2002 22:37:37:
Als Antwort auf: / In reply to: Re: Wb2Uci and King geschrieben von: / posted by: Thomas Lagershausen at 23 September 2002 21:04:06:

First thanks a lot for this amazing tool.
It makes that i can analyze with this fantastic king-engine under Fritz and the shredder-gui.
Now for my problem.
With all versions of your adapter the king-engine stops always(!) thinking in analyze-mode after between about 44 and 72 minutes(hardware:Athlon 1,5 Ghz).
I am a tournament player and love it to let two programs run the whole night on a position.
It would be a dream if you can find the problem.
With the inbetween ini it works, but the WB2Uci tool is off course better.
Greetings
Thomas Lagershausen

Hmm.
I use the analyze function of the engine for 'go infinity' that is the analysis in Fritz.
I will check it up.
There is a switch for it but it is on as default.
Analyze=true
To simulate the analyze function for engines without this build in you have to set it to false. Then I have seen that it could stop.
I guess I have never had time to test it for so long time
I set it on tonigh so I can read the log-file tomorrow.
Wich version? I test the newest (The King 3.23 or CM9000).
Odd Gunnar
Odd Gunnar

[Andreas Herrmann]

Geschrieben von: / Posted by: Andreas Herrmann at 23 September 2002 22:47:42:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Odd Gunnar Malin at 23 September 2002 22:29:47:

Hi Odd,
please, test Yace and implement time control " a'/b + c'/d + e' ".
Thanks,
/Matthias.

Is there any problem with the uci version?
Anyway I did a run with timecontrol 1'/25 + 1'/15 +1'/g just now and the adapterversion won 4-0 against the uci version.

The last time I tested it under Fritz7 GUI at 10'/40+5'/20+6',
engines (Holmes+Cerebro) would play normally during first 40 moves and after move 60, but use only about 0 seconds in 2nd time interval (moves 41 ... 60).
I will test this behaviour again.
/Matthias.

For Holmes you need LevelExtend=strict
Ex.
[ENGINE]
Name=Holmes 0.80
Author=Andreas Herrmann
Filename=Wb2Uci.exe
[OPTIONS]
Program=Holmes080.exe xboard
OwnBook=true
Edit=edit
LevelExtend=strict

Maybe I was abit fast here, when I try with exactly your timecontrol I have to use LevelExtend=failsafe Eg. on t/40+t/20+t/g
I will study this a bit more.
Odd Gunnar

Hi Odd Gunnar,
is it possible to use wb2uci under Shredder 5? That's the only professional GUI that seems using the UCI protocoll, i have. Because i want to check your wb2uci by myself.
Andreas

Yes I would think so. I have been using it under Shredder 6 (Classic) and the april version (1.2) is tested for it. I have not come so long in my testing on the new version that I have tested it in Shredder.
I guess I expect it to run smoothly in there so I leave it to the final test 1 minute before I release it (remove the beta).
You could also try it on Arena. There you can match it against itself as a wb-engine to see any differences.
Odd Gunnar

Yes good idea to test it under Arena against itself as a wb.
have a nice day
Andreas



Homepage of Holmes

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 23 September 2002 23:10:49:
Als Antwort auf: / In reply to: Re: Wb2Uci geschrieben von: / posted by: Odd Gunnar Malin at 23 September 2002 19:38:06:

[ENGINE]
Name=Holmes 0.80
Author=Andreas Herrmann
Filename=Wb2Uci.exe
[OPTIONS]
Program=Holmes080.exe xboard
OwnBook=true
Edit=edit
LevelExtend=strict

Maybe I was abit fast here, when I try with exactly your timecontrol I have to use LevelExtend=failsafe Eg. on t/40+t/20+t/g
I will study this a bit more.
Odd Gunnar

Hi Odd,
Cerebro now works fine !! Only Holmes will not obey yet.
Maybe I was too "strict" with him. Now I shall try "failsafe".
Thanx,
Matthias.

[Thomas Lagershausen]

Geschrieben von: / Posted by: Thomas Lagershausen at 23 September 2002 23:11:27:
Als Antwort auf: / In reply to: Re: Wb2Uci and King geschrieben von: / posted by: Odd Gunnar Malin at 23 September 2002 22:37:37:

First thanks a lot for this amazing tool.
It makes that i can analyze with this fantastic king-engine under Fritz and the shredder-gui.
Now for my problem.
With all versions of your adapter the king-engine stops always(!) thinking in analyze-mode after between about 44 and 72 minutes(hardware:Athlon 1,5 Ghz).
I am a tournament player and love it to let two programs run the whole night on a position.
It would be a dream if you can find the problem.
With the inbetween ini it works, but the WB2Uci tool is off course better.
Greetings
Thomas Lagershausen

Hmm.
I use the analyze function of the engine for 'go infinity' that is the analysis in Fritz.
I will check it up.
There is a switch for it but it is on as default.
Analyze=true
To simulate the analyze function for engines without this build in you have to set it to false. Then I have seen that it could stop.
I guess I have never had time to test it for so long time
I set it on tonigh so I can read the log-file tomorrow.
Wich version? I test the newest (The King 3.23 or CM9000).
Odd Gunnar
Odd Gunnar

It´s fantastic that you notice my problem.
The case is with both versions of king (3.12 and 3.23)
Here is my eng:
[ENGINE]
Name=The King 3.23
Author=Johan de Koning, NED
Filename=Wb2Uci.exe
[OPTIONS]
Program=TheKing.exe
InitString=/xboard := xboard\npost\nnew\ncm_parm default\ncm_parm opk=421666 tts=33554432
HashCommand=hash %iM
Hash=32
HashOnCommandline=false
InitTime=2
Edit=setboard
UseSAN=false
Ponder=false
Logfile=false
Priority=Normal
Protocol=2
ShowThinkingMove=true
Thanks for all.
Thomas L.

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 23 September 2002 23:17:57:
Als Antwort auf: / In reply to: Thanks !! geschrieben von: / posted by: Matthias Gemuh at 23 September 2002 23:10:49:

Hi Odd,
Cerebro now works fine !! Only Holmes will not obey yet.
Maybe I was too "strict" with him. Now I shall try "failsafe".
Thanx,
Matthias.

Failsafing Holmes does not help either.
BTW, Cerebro is private and in Leo's Division 2.
/M.

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 24 September 2002 00:20:29:
Als Antwort auf: / In reply to: Holmes geschrieben von: / posted by: Matthias Gemuh at 23 September 2002 23:17:57:

Hi Odd,
Cerebro now works fine !! Only Holmes will not obey yet.
Maybe I was too "strict" with him. Now I shall try "failsafe".
Thanx,
Matthias.

Failsafing Holmes does not help either.
BTW, Cerebro is private and in Leo's Division 2.
/M.

Strange, failsafe you work for all engines that have force command.
That the adapter do is when it comes to a timecontrol eg. at move 40 and 60 it will send all previous move as if it was bookmove and then assign the new time control. Ex.
level 10 40 0
// play 40 moves
level 5 60 0
// feed the 40 previous moves as it was book moves.
// play 20 moves
level 0 6 0
// feed the 60 previous moves as it was book moves.
// play the rest of the game.

This was the method the old CB adapter did with the exeption that it did it for every moves and not only two times in a game.
The other two modes (progressive and strict) try to set the level without sending the moves, the difference is how it calculate(*) the level command.
Well I have to test it a bit more then.
If it happend every game on your machine, could you enable logging and send me the logfile.
Odd Gunnar
(* As a footnote without beeing involved. Ruffian do this calculation different from Crafty, TheKing, Gandalf and Yace.)

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 24 September 2002 01:40:20:
Als Antwort auf: / In reply to: Re: Holmes geschrieben von: / posted by: Odd Gunnar Malin at 24 September 2002 00:20:29:

Strange, failsafe you work for all engines that have force command.
That the adapter do is when it comes to a timecontrol eg. at move 40 and 60 it will send all previous move as if it was bookmove and then assign the new time control. Ex.
level 10 40 0
// play 40 moves
level 5 60 0
// feed the 40 previous moves as it was book moves.
// play 20 moves
level 0 6 0
// feed the 60 previous moves as it was book moves.
// play the rest of the game.

This was the method the old CB adapter did with the exeption that it did it for every moves and not only two times in a game.
The other two modes (progressive and strict) try to set the level without sending the moves, the difference is how it calculate(*) the level command.
Well I have to test it a bit more then.
If it happend every game on your machine, could you enable logging and send me the logfile.
Odd Gunnar
(* As a footnote without beeing involved. Ruffian do this calculation different from Crafty, TheKing, Gandalf and Yace.)

I think the problem is in Holmes, not wb2uci.
Andreas told me he was not supporting CB protocol (in Holmes 0.8.0 or 0.8.2).
I don't have password to unzip and test Holmes 0.8.4.
Regards,
Matthias.

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 24 September 2002 01:54:02:
Als Antwort auf: / In reply to: Re: Holmes geschrieben von: / posted by: Matthias Gemuh at 24 September 2002 01:40:20:

Strange, failsafe you work for all engines that have force command.
That the adapter do is when it comes to a timecontrol eg. at move 40 and 60 it will send all previous move as if it was bookmove and then assign the new time control. Ex.
level 10 40 0
// play 40 moves
level 5 60 0
// feed the 40 previous moves as it was book moves.
// play 20 moves
level 0 6 0
// feed the 60 previous moves as it was book moves.
// play the rest of the game.

This was the method the old CB adapter did with the exeption that it did it for every moves and not only two times in a game.
The other two modes (progressive and strict) try to set the level without sending the moves, the difference is how it calculate(*) the level command.
Well I have to test it a bit more then.
If it happend every game on your machine, could you enable logging and send me the logfile.
Odd Gunnar
(* As a footnote without beeing involved. Ruffian do this calculation different from Crafty, TheKing, Gandalf and Yace.)

I think the problem is in Holmes, not wb2uci.
Andreas told me he was not supporting CB protocol (in Holmes 0.8.0 or 0.8.2).
I don't have password to unzip and test Holmes 0.8.4.
Regards,
Matthias.

No, this have little or nothing to do with the Fritz protocol.
It will send the the moves as Chessmaster/Chesspartner/ChessAssistant etc. when you use their internal book.
It will clear the hash two times in the game so on these two moves it will use a little more time. But the reward is that more people will play human-comp games against it.
Odd Gunnar
(My test pc is busy with testing TheKing in analyse mode so I don't get time until tomorrow to do more tests on this.)

[Andreas Herrmann]

Geschrieben von: / Posted by: Andreas Herrmann at 24 September 2002 02:24:49:
Als Antwort auf: / In reply to: Re: Holmes geschrieben von: / posted by: Matthias Gemuh at 24 September 2002 01:40:20:

Strange, failsafe you work for all engines that have force command.
That the adapter do is when it comes to a timecontrol eg. at move 40 and 60 it will send all previous move as if it was bookmove and then assign the new time control. Ex.
level 10 40 0
// play 40 moves
level 5 60 0
// feed the 40 previous moves as it was book moves.
// play 20 moves
level 0 6 0
// feed the 60 previous moves as it was book moves.
// play the rest of the game.

This was the method the old CB adapter did with the exeption that it did it for every moves and not only two times in a game.
The other two modes (progressive and strict) try to set the level without sending the moves, the difference is how it calculate(*) the level command.
Well I have to test it a bit more then.
If it happend every game on your machine, could you enable logging and send me the logfile.
Odd Gunnar
(* As a footnote without beeing involved. Ruffian do this calculation different from Crafty, TheKing, Gandalf and Yace.)

I think the problem is in Holmes, not wb2uci.
Andreas told me he was not supporting CB protocol (in Holmes 0.8.0 or 0.8.2).
I don't have password to unzip and test Holmes 0.8.4.
Regards,
Matthias.

Hi Matthias,
i have just send you the passwords. It would be nice if you can send me a logfile of Holmes. Perhaps there is an initialisation problem in Holmes. I have written the time code where the wb1 protocoll was out and there was normally no way to change a level during a game. So perhaps i have to rewrite parts of the code.
have a nice day
Andreas

[Odd Gunnar Malin]

Geschrieben von: / Posted by: Odd Gunnar Malin at 24 September 2002 09:25:08:
Als Antwort auf: / In reply to: Re: Wb2Uci and King geschrieben von: / posted by: Thomas Lagershausen at 23 September 2002 23:11:27:

First thanks a lot for this amazing tool.
It makes that i can analyze with this fantastic king-engine under Fritz and the shredder-gui.
Now for my problem.
With all versions of your adapter the king-engine stops always(!) thinking in analyze-mode after between about 44 and 72 minutes(hardware:Athlon 1,5 Ghz).
I am a tournament player and love it to let two programs run the whole night on a position.
It would be a dream if you can find the problem.
With the inbetween ini it works, but the WB2Uci tool is off course better.
Greetings
Thomas Lagershausen

Hmm.
I use the analyze function of the engine for 'go infinity' that is the analysis in Fritz.
I will check it up.
There is a switch for it but it is on as default.
Analyze=true
To simulate the analyze function for engines without this build in you have to set it to false. Then I have seen that it could stop.
I guess I have never had time to test it for so long time
I set it on tonigh so I can read the log-file tomorrow.
Wich version? I test the newest (The King 3.23 or CM9000).
Odd Gunnar
Odd Gunnar

It´s fantastic that you notice my problem.
The case is with both versions of king (3.12 and 3.23)
Here is my eng:
[ENGINE]
Name=The King 3.23
Author=Johan de Koning, NED
Filename=Wb2Uci.exe
[OPTIONS]
Program=TheKing.exe
InitString=/xboard := xboard\npost\nnew\ncm_parm default\ncm_parm opk=421666 tts=33554432
HashCommand=hash %iM
Hash=32
HashOnCommandline=false
InitTime=2
Edit=setboard
UseSAN=false
Ponder=false
Logfile=false
Priority=Normal
Protocol=2
ShowThinkingMove=true
Thanks for all.
Thomas L.

It did not stop during the night here. I used a bit easier eng file but only yuor initstring and hashcommand is somewhat wrong but it should not result in
this stop.
[ENGINE]
Name=The King 3.23
Author=Johan de Koning
Filename=Wb2Uci.exe
[OPTIONS]
Program=TheKing3.23.exe
InitString=cm_parm default\ncm_parm opk=xxx
HashCommand=cm_parm tts=%i004096
Hash=1
Visible=Ponder,Hash,ShowThinkingMove,Logfile
Then I set it up inside Fritz to use 16MB hashtable size.
I will see when I get home if it still running.
I checked the logfile ind it use the right 'analyze' function.
Odd Gunnar