Page 1 of 3

WinBoard 4.7.3 released

Posted: Tue Jan 07, 2014 9:52 pm
by hgm
Two days ago the source code of XBoard 4.7.3 was released. I just uploaded a binary installer package for the corresponding WinBoard version to my website: .

This is just one of the bugfix releases we regularly do of the stable branch, so don't expect any new features. There wasn't actually much to fix; the complete list of patches compared to 4.7.2 is this:

Code: Select all

    Fix buffer overflow in parser on long PGN with scores
    Allow castling and e.p. capture moves to survive when editing opening book
    Fix node-count overflow in engine output
    Better sorting of fail-low lines in engine output
    Fix Chess960 a-side castling legality check
    Fix position search in Chess960
    Fix adjudication of Giveaway stalemates
    Adapt WinBoard window positioning to work with multiple screens
    Cure some weirdness in XBoard when dragging pieces outside the board
    Some dialog layout improvements to facilitate internationalization
    Fix crash on using some dialog Browse buttons in GTK version
    Connect manual menu item work in GTK version
    Put code for OS X integration in the GTK version
The last 5 fixes only affect XBoard. Some of the bugs in variants could be pretty annoying, but of course would not impact normal Chess. The book editing is really the only important fix as far as orthodox Chess is concerned. This release was mainly driven by fixes in XBoard, which included an important bug in the GTK build, crippling most Browse buttons in dialogs. Another significant development there was porting the GTK version for Mac from using XQuartz to the more native Quarz API, with as a result that XBoard is now integrated into OS X as if it was a native application. (Thanks to the work of Joshua Pettus!)

For the adventurous, there also is a new release of the development version available; the WinBoard binary for this can be downloaded from , and could be used to replace the winboard.exe in the 4.7.3 install. This version does contain several new features. What would be of interest for orthodox Chess is

* New options -first/secondDrawDepth N
for speeding up play of tablebase-using engines (by limiting their search depth to N) in positions that according to a bitbase probe are a draw. This would prevent CPU wastage by engines trying to swindle each other in a drawn situation, by forcing them to instantly play the drawing move which according to their evaluation and a very shallow search is best. (In won and lost positions tablebase-using engines should already move instantly by themselves, without the help of any options.)

For these options to work, bitbases should be installed, and the -egtFormats string should contain the element scorpio:BITBASEPATH (can be typed in the Nalimov Path textedit of the Common Engine dialog)

* Edit Position mode was improved
Repeatedly exercising 'clear' will now cycle the board state through empty - pallette - initial position - original position - empty ... Where 'original position' is the position that you last completely erased, so that it is possible to recover if you cleared the board by mistake. The 'pallette' position contains all non-Pawns exactly once, and is a very convenient way to set up positions with just a few pieces, where you can move the required once to the desired squares, possibly duplicating one by moving with Ctrl pressed, or throwing it off board (or capturing it) when not needed. (Pawns can be added with a static mouse (right-)click, which is even easier than moving them, white for a plain click, black when Shift pressed.)

The variants ASEAN (a new form of Makruk) and Mighty-Lion Chess were added (the latter as a spin-off of Chu Shogi, which was added on XBoard). It also allows engines to define variants with non-standard names, which will appear in the New Variant dialog as choices. When the user selects such an engine-defined variant, the engine must tell WinBoard the specifics (board size, initial position, parent variant, participating pieces) through a 'setup' command, now officially part of WB protocol. (It existed before as undocumented feature.) This makes it, for instance, possible to make engines for all the ICS 'wild' boards, without the need for the user to configure WinBoard for them by providing initial positions from a file, etc.

Have fun, have WinBoard!

Re: WinBoard 4.7.3 released

Posted: Tue Jan 07, 2014 10:24 pm
by Daniel Shawul

Re: WinBoard 4.7.3 released

Posted: Tue Jan 07, 2014 10:39 pm
by hgm

Re: WinBoard 4.7.3 released

Posted: Wed Jan 08, 2014 12:12 am
by Frank Quisinsky
Works fine now!
Thanks, I have to look :-)


Re: WinBoard 4.7.3 released

Posted: Wed Jan 08, 2014 1:05 am
by Kyodai

Thanks HG!

In WinBoard I use a rather big polyglot opening book (.bin)

about 413MB in gross. Now, I wonder how to add new games

to this book - using Win 7 ??

Re: WinBoard 4.7.3 released

Posted: Wed Jan 08, 2014 8:28 am
by hgm
Polyglot can be run as stand-alone program for converting PGN files for orthodox Chess to book, and for merging multiple books into one. I have never done the latter, however, and the former only once, and don't remember the details. You just run Polyglot from the command line with some flag argumments and filenames.

This is probably the easiest way to do it. WinBoard can convert PGN files to book through the 'Save games as book' item, but not merge existing books. Polyglot probably creates book from PGN faster anyway; the WinBoard feature is mainly intended for variants Polyglot cannot handle. WinBoard can also be used to edit an existing book, for purpose of tuning: change the weight of moves in specific positions, add or delete moves. This is something Polyglot cannot do. But this is only important for people that make books by hand, rather than by taking statistics from game datasets.

Re: WinBoard 4.7.3 released

Posted: Wed Jan 08, 2014 6:35 pm
by mohzus
H.G.Muller wrote:Fix node-count overflow in engine output

I'm afraid to report that the nps displayed in the Engine Output is still not fixed in xboard 4.7.3. I just compiled and installed it (xboard -v returns the 4.7.3 version in the terminal).
I still get a "cut" nps as with the previous versions.
Screenshot attached.
As you can see, a single digit is missing.

With respect to the "sorting of fail-low lines in engine output" I do not know what it really is, but the most recently calculated line is not always on the top of the list of lines, in the Engine Output. Is this normal?

Re: WinBoard 4.7.3 released

Posted: Wed Jan 08, 2014 7:13 pm
by hgm
Seems I goofed again. I made this fix only in the development version, and it did not find its way into v4.7.x. :cry:

Multiple lines of the same depth are sorted by score. This to give better display in multi-PV mode. As engines would normally only supercede a line with a new one if it has higher score, this should not affect single-PV mode.

The problem, however, is that Stockfish sends fail-low scores disguised as real PVs. Such a score is just an upper limit, and triggers a re-search, which should find the real score, which will be lower. I improved the sorting now to sort lines for the same move always according to the order in which they were received, as the same move can only get another score at the same depth if the first score was invalid. It then internally corrects the score of the older moves to the latest one, so that sorting w.r.t. moves coming later will be done correctly.

I thought this would solve it. (It did for the cases I tried.) But if this is still not satisfactory, I guess that it can only be solved in Polyglot. There the information whether the score was a fail low might be directly available.

Re: WinBoard 4.7.3 released

Posted: Wed Jan 08, 2014 7:14 pm
by hgm
I now included a French language file in the install, thanks to Ben St-Pierre! (Same link.)

Re: WinBoard 4.7.3 released

Posted: Wed Jan 08, 2014 7:55 pm
by mohzus
No problem H.G. Muller. I have edited the nengineoutput.c to fix the problem as you suggested in a thread in winboard forum. It works great now.

I have a question about the clocks size. I do not know how to reduce their size, is there a way I could do this?
By the way I used the Xaw compile and not the GTK one.