WinBoard 4.3.15 released

[slobo]

Anyway, I think it is preferable to fix this problem in SCID rather than bend the standard by redefining of the weights.

It is not a problem with SCID. It is a limitation of the Polyglot Book format.

If PG refuses to make books with zero weights and even has an ASSERT in its probing routine to verify this I don't think SCID should start generating books with weights that are mathematically zero.

It is also not bending the standard since currently there is no standard how the weights are to be interpreted. So we can make the standard whatever we want to. So stating
that moves with probability < 0.001 should be ignored is perfectly valid since it changes current behaviour only in 0.1% of the cases and fixes an ambiguity.

But it was just a suggestion. If you don't like it, just ignore it.

Why do not adopt Arena´s solution with priorities, numbers from 0 to 9?

[Michel]

Why do not adopt Arena´s solution with priorities, numbers from 0 to 9?

Changing the interpretation of the weights would break all current books.

[slobo]

The /ini argument asks WB to save the settings in that file (on exit or when you press the menu).

If you don't want that, you should use @wood.ini in the target field of the shortcut.

To use the polyglot book, go to the Options -> UCI... menu. Make sure the Polyglot Book field is set to the filename of a valid book, and tick "use Book". Then untick the boxes "hasOwn Book" for the engines that you want to use the GUI book. Finally click "OK". That should do it.

I changed the names of my 4 small and addicional ini files: they all now star with @ character, as you suggested. As a consequence, when I Save Setting Now, from the Option menu, my small and additional ini file is not replaced by winboard.ini. So, one problem is solved. But there is another problem: Save Setting Now does not work anymore. I have to edit winboard.ini to change things.

Is there any way to use a small additional ini file from the command line without the /ini argument?

regards

Perhaps the best solution would be to save WB configuration changes always in winboard.ini, and not in an ini file with /ini argument in the command line?

[hgm]

I changed the names of my 4 small and addicional ini files: they all now star with @ character, as you suggested. As a consequence, when I Save Setting Now, from the Option menu, my small and additional ini file is not replaced by winboard.ini. So, one problem is solved. But there is another problem: Save Setting Now does not work anymore. I have to edit winboard.ini to change things.

Is there any way to use a small additional ini file from the command line without the /ini argument?

The names should not start with '@'. The way to use additional files from the command line is to prefix the filename with '@'.

So if you put @xxx amongst the arguments on the command line, WB will read the file with name xxx, and inserts its contents into the command line at that point (replacing any newline in the file by a space).

The /ini argument is not needed for all this: it is an argument mainly for specifying where the settings should be saved.

[hgm]

It is not a problem with SCID. It is a limitation of the Polyglot Book format.

If PG refuses to make books with zero weights and even has an ASSERT in its probing routine to verify this I don't think SCID should start generating books with weights that are mathematically zero.

I do consider this a SCID problem: it should delete the entry for moves it does not want to be played. This is apparently the way the Polyglot standard specifies it should be done. Fiddling with the weights is simply a shortcut invented by SCID, which is not guaranteed to work. What, for instance, if you want to suppress the only move in the book from a given position? Or if there are two moves, each with weight 1?

It is also not bending the standard since currently there is no standard how the weights are to be interpreted. So we can make the standard whatever we want to. So stating
that moves with probability < 0.001 should be ignored is perfectly valid since it changes current behaviour only in 0.1% of the cases and fixes an ambiguity.

But it was just a suggestion. If you don't like it, just ignore it.

Indeed I don't like it. The standard was implicitly defined by how Polyglot uses the book. And if Polyglot used to play moves with a probability of less than 0.001, having other book users play those moves with a different probability, makes the same book behave differently from one usere to the other. I think that in principle that is a very bad idea.

You can argue that a difference of 0.001 in playing probability is not significant in practice. But that works the other way around as well; those that want 0 should be able to live with 0.001. Those that want 0.001 should be able to get it. The former group can get the 0 by deleting the move from the book, as SCID should have done in the first place. It is not clear what the alternative should be for the second group...

[Michel]

What, for instance, if you want to suppress the only move in the book from a given position? Or if there are two moves,

Well this is unlikely to occur in book tuning. You usually want to help the engine picking a particular move, not leave it on its own.

Or if there are two moves, each with weight 1?

This I don't understand. If you give one of the moves zero weight the other one will be played 100% of the time.

Currently to simulate zero weight Scid sets one of the weights equal to 1 and the other to 65535.

[hgm]

OK, so it even changes the weights of moves that you don't want to touch.

Anyway, removing the entry seems the proper way to solve the problem. But if the weights can run upto 2^16 the probability can be set to 0.001%, which seems low enough to not worry about it.

[slobo]

Another question:
Where could I see the menue that reflects the following lines from winborad.ini:

/icsMenu={-
&Who,who,0,1
Playe&rs,players,0,1
&Games,games,0,1
&Sought,sought,0,1
|&Tell (name),tell,1,0
M&essage (name),message,1,0
-
&Finger (name),finger,1,1
&Vars (name),vars,1,1
&Observe (name),observe,1,1
&Match (name),match,1,1
Pl&ay (name),play,1,1
}

I tried to find it, but in vain.

Second.

I entered FICS as a guest, start a game [play (nimber)], but my moves as black, in two games, were always taken back. WB did not allow me to play moves.

[Michel]

OK, so it even changes the weights of moves that you don't want to touch.

I don't see why that's a problem....

The weights only have a relative meaning.

[hgm]

The string seems indeed not used inside WinBoard. No idea who added it and why. It seems a WinBoard menu definition, but it does not seem very sensible to define menus through command-line options.

It is difficult to say anything about the FICS problem, without seeing a debug file. E.g. is it sure that it is WinBoard that refuses the move, or was it FICS? Usually WinBoard gives a reason when it refuses a move, in the message area above the board. (Like: "It is black's turn" or "Illegal move".)