Bunding WinBoard and suport programs

[H.G.Muller]

What about the following file-system tree for the bundled release of WinBoard plus support programs (let's call it "WinBoard Gold Pack"):

Code: Select all

Chess\
    README.TXT
    WinBoard Gold.html         (main page of html manual)
    WinBoard\
        winboard.exe
        winboard.ini
        WINBOARD.hlp
        Fancy Look.lnk             (Sample shortcuts, starts CPW + wooden board)
        Capablanca Chess.lnk  (Sample shortcut for Fairy-Max in variant)
        fancy.ini
        capa.ini
        Textures\
            Wood\
                dave_dark.bmp    (Needs permission from David Dahlem)
                dave_lite.bmp
    Polyglot14w\
        polyglot.exe
        book.dat                 (Is there a small Polyglot book we can include?)
        CPW.ini
        Polyglot.html           (UCI engine installation manual)
    PSWBTM1\
        PSWBTM.exe
        config.pswbtm
        engine.pswbtm
        ntls.pswbtm
        winboard.ini
        Nunn.pgn
        SilverSuite.fen               (Needs permission of Albert Silver)
        PSWBTM.html                (PSWBTM manual)
        games\
    Fairy-Max\
        fmax.exe
        fmax.ini
        CVfairy.html
        fairysmall.jpg
        readme.txt
    CPW\
        cpw.exe
    SMIRF
        Smirfoglot.exe
        readme.txt


[Guenther Simon]

What about the following file-system tree for the bundled release of WinBoard plus support programs (let's call it "WinBoard Gold Pack"):

Code: Select all

        book.dat                 (Is there a small Polyglot book we can include?)
...


How small is small? I would offer my medium book which is permanently
updated or even a smaller one if I know how small it should be especially
in ply length
(Same for a PSWBTM engines file.)

BTW what about asking Tommy too? No idea though if TLCV or Andrews
equivalents and Tommys GameAnalyzer are more for an advanced package?
At least I would say LGPGNVER should be a must if George would agree...

Regards,
Guenther

[Roger Brown]

How small is small? I would offer my medium book which is permanently
updated or even a smaller one if I know how small it should be especially
in ply length
(Same for a PSWBTM engines file.)

BTW what about asking Tommy too? No idea though if TLCV or Andrews
equivalents and Tommys GameAnalyzer are more for an advanced package?
At least I would say LGPGNVER should be a must if George would agree...

Regards,
Guenther

Hello Guenther,

This idea is getting better by the day.

Of course I was going to get around to asking you for something and here you are volunteering. This is splendid!

The book is a definite yes and so is the PSWBTM engines file.

I hope that H.G. will get back to you but if he does not, work your wizardry and produce the best small book you can and I will push it.

Well, if we are going gold let us go the whole way I say. I tried to contact Tommy some time ago and there was no response. Could you ask him about his Winboard ini file gui and his viewer for me?

As you would have better contacts than me with George, could I ask you to ask him? I hope he agrees!

H.G. seems to be inferring different packages from the simple "I just want to start to start now " to a "max pack" bundle with extras.

Hmmmm, this is a good development.

Later.

[H.G.Muller]

As to the book size: the only thing I want to prevent is that anyone component of the package starts to dominate it, size-wise. The components outside the essential software are just meant as examples of how to do things, but still offer enough incentive for people to become active and replace them by someting better that they conciously choose. This is why I don't want to include, say, Toga II and the most elaborate book in existence. It would make people lazy, and it would be unfair to the alternatives.

I am not sure if this is a stupid idea, or not; this is just meant as a discussion proposal, you don't have to take my word for anything. I am sure there are people on this forum much more qualified than I am in designing a good WinBoard operating environment.

As to th PSWBTM engine file: I suppose Guenther's engine.pswbtm contains invaluable information on the best options for installing engines. OTOH, I am not sure if providing users with a PSWBTM with a completely filled engine database would be useful if they don't have the engines themselves already on their system. They might download them in other folders than the database assumed. And engines are likely to be replaced with newer versions over time, wich might have different names for their executables or folders in which they download. There is a lot of room there to create confusion.

Would it be an idea to provide them with an engine.pswbtm that does contain entries for all engines, with a name and a the option settings, but no path for executable yet. Or, if PSWBTM doesn't allow that, put a name of a dummy engine there, which we call "Yet to be downloaded". So that people are reminded that they still have to do this. We could even provide a "Yet to be downloaded.exe" pseudo-engine, which, when invoked by WinBoard, would pop up a screen pointing out the user selected an engine he doesn't have yet, with instructions how to download engines and install them in PSWBTM. (Or, more modestly, has WB pop up an error-mesage box ointing the user to a certain manual file, using the "telluser" command.)

About George Lyapko's tools: I have little experience in using them, but I understand they are in general extremely useful. If George has no objection, we could add them in a Tools/ folder. But specifically about PGNVER: is this still useful when running WinBoard 4.3? I have a program similar to George's LGTBL, which produces a cross-table from a PGN file not in HTML but as ASCII text, (e.g. useful for posting it in a forum). We could include that too.

I have my doubts about the broadcasting stuff (TLCS). I am not sure many people would want to do that. OTOH, if it is not too large, it would not hurt anyone including it. I have never used it, so I have no idea how involved it is to set up a broadcast. (Is it just TLCS, or do you need other server packages to allow it to run?) It is not possile for me to contact Tommy. I have tried it many times, but apparently my mails do not make it through his spam filter.

[Ron Murawski]

About George Lyapko's tools: I have little experience in using them, but I understand they are in general extremely useful. If George has no objection, we could add them in a Tools/ folder. But specifically about PGNVER: is this still useful when running WinBoard 4.3? I have a program similar to George's LGTBL, which produces a cross-table from a PGN file not in HTML but as ASCII text, (e.g. useful for posting it in a forum). We could include that too.

George Lyapko's tools don't seem to run correctly on my 64-bit system.

Ron

[Volker Pittlik]

[quote="Ron Murawski"...

George Lyapko's tools don't seem to run correctly on my 64-bit system.

Ron[/quote]

Something similar happens with wine on 64 bit Linux. IIRC wine complains that it can't run dos programs. The tools seem to be 16 -bit.

vp

[Thomas McBurney]

Hi Everyone!

I'm surprised that there was so many people trying to contact me. I guess the Yahoo spam filter is pretty good at filtering emails that you want to receive.

Anyway, as for the broadcast server and Game Analyser. I don't mind if they are included in this proposed package. The broadcast server is 130Kb in size and does not rely on any special external components. In fact, this may encourage some people to have a go at broadcasting their tournaments. The only thing about the broadcast server is that it is a bit difficult for people to get working because they need to understand how to configure their firewalls, NAT Routers etc, so that people can connect from the outside.

The Game Analyser I have recently updated so that it can now interpret PGN's from Arena. I have also made some visual changes but I have not uploaded it to my web site yet.

About 2.5 years ago, I was working with Alessandro and Roger on integrating my engine manager with Winboard. We actually got this working before we lost interest. Many of the features we planned to include for this project were not completed, such as the ability to add UCI engines from the engine manager. Here some screen shots...



After clicking on 'Manage'...



If anyone is interested in the newer version of Engine Manager (which is actually 2.5 years old) then let me know. I don't have enough time to work on it so I'm interested in releasing the source if someone wants to develop it further to include UCI and other features.

Cheers,
Tom.

[Roger Brown]

Hi Everyone!

I'm surprised that there was so many people trying to contact me. I guess the Yahoo spam filter is pretty good at filtering emails that you want to receive.

Hello Thomas,

It is very good to hear from you! That spam filter comment had me laughing.

Anyway, as for the broadcast server and Game Analyser. I don't mind if they are included in this proposed package.

Thanks much kind sir.

[SNIP]

The Game Analyser I have recently updated so that it can now interpret PGN's from Arena. I have also made some visual changes but I have not uploaded it to my web site yet.

Ahem. Why not?

About 2.5 years ago, I was working with Alessandro and Roger on integrating my engine manager with Winboard. We actually got this working before we lost interest.

Good times, good times. I enjoyed that experience.

Many of the features we planned to include for this project were not completed, such as the ability to add UCI engines from the engine manager.

And then there was Polyglot so even though it is not direct the job gets done.

If anyone is interested in the newer version of Engine Manager (which is actually 2.5 years old) then let me know. I don't have enough time to work on it so I'm interested in releasing the source if someone wants to develop it further to include UCI and other features.

This is indeed gracious of you. Thanks for making your intellectual property freely available.

You mention that you have little time for such vital pursuits as the Engine Manager. Does that include your chess engine?

All the best sir.

Later.

[H.G.Muller]

Is TLCS open source? I am not sure we can distribute software with WinBoard of which we are not prepared to hand over the source, or we might be violating the GPL on WinBoard.

About the engine manager: Is this part of WinBoard, or is it a separate program? From the screenshots it seems like you start it from the WinBoard menu. What exactly does it do? Is it an editor for the winboard.ini file, which upates the "/chessProgramNames" options in it?

I am not sure how useful this wil be, especially if it is not finished. Since I have PSWBTM I never install any engines in the winboard.ini file whatsoever; this is all handled by PSWBTM, wich maintains its own engine database. I'd like to hear input from others about this; I think we should be careful not to overload the user with many alternatives to do the same thing. That will be as detrimental as giving them too little software. They will then have everything they need in one download, but won't be able to find the components they most commonly need amongst everything they have, and that will deter them as well. We have to strike a delicate balance, there.

Now that Thomas is here: One of the reasons I wanted to contact you before is this: some engines seem to crash TLCS by overloading the winboard.debug file with massive protocol-violating output. WinBoard ignores this, but it does end up in the debug file. Apparently it then overruns some input buffers in TLCS. Or perhaps they contain some keywords that confuse TLCS.

Would it be easy for you to change TLCS much that it ignores any line that starts with a '#' sign? I have included an option to WinBoard that would prefix every protocol-violating input line it ignores with '#' before writing it into the .debug file, for engines that do not do this themselves. I would be surprised if having TLCS simply ignore such lines would not solve almost all problems with such rogue engines.

[Roger Brown]

Is TLCS open source? I am not sure we can distribute software with WinBoard of which we are not prepared to hand over the source, or we might be violating the GPL on WinBoard.

Hello H.G.

No, it is not open source.

About the engine manager: Is this part of WinBoard, or is it a separate program? From the screenshots it seems like you start it from the WinBoard menu. What exactly does it do? Is it an editor for the winboard.ini file, which upates the "/chessProgramNames" options in it?

It is a separate program and yes, it updates the Winboard ini file.

I am not sure how useful this wil be, especially if it is not finished. Since I have PSWBTM I never install any engines in the winboard.ini file whatsoever; this is all handled by PSWBTM, wich maintains its own engine database. I'd like to hear input from others about this; I think we should be careful not to overload the user with many alternatives to do the same thing. That will be as detrimental as giving them too little software. They will then have everything they need in one download, but won't be able to find the components they most commonly need amongst everything they have, and that will deter them as well. We have to strike a delicate balance, there.

I guess I disagree here. It depends on what the primary purpose is. It may be that a user might not want to bother with tournaments initially. When the bug bites then they might want to get into gauntlets, round robins etc.

The good thing here is that there are is compatibility as Winboard ini can be imported by PSWBTM (indeed by all the major free WBTM's, I made sure of that! ) if desired.

I can see where a new user might feel ooverwhelmed with all the choices but I believe they will make it if the packages are appropriately assembled.

Later.

[H.G.Muller]

So if it is not open source, or indeed not GPL, I foresee legal problems.

If the engine manager is a separate program, what is the licencing status of that one? It seems that to behave as in the screenshot, it would need some modifications in WinBoard. Currently, there is no "manage..." in the Options menu.

Btw, I do not use PSWBTM only for tournaments. If I would want to run an engine for playing human-engine (or as human-operated engine) I do it through PSWBTM. It has the "launch engine" button for that in the engine-manager dialog.

So you can import engines into PSWBTM from the winboard.ini? I never new that! So this is what the "Import" buttons in the engine manager are for? But there is also an "Export" button. Doesn't this already do what you want?

[Roger Brown]

[SNIP]

Btw, I do not use PSWBTM only for tournaments. If I would want to run an engine for playing human-engine (or as human-operated engine) I do it through PSWBTM. It has the "launch engine" button for that in the engine-manager dialog.

I just start up Winboard, select from the drop down menu and go. Now that I think about it though, the added advantage in using the TM is that I can change the opponent much quicker and easier. Besides PSWBTM saves me the annoyance of starting Winboard and waiting for that annoying UAC and startup dialogue box so I guess you may very well have a point here.

Hmmmmm, the TM is easier to use than the program directly. Thanks for the tip H.G.

So you can import engines into PSWBTM from the winboard.ini? I never new that! So this is what the "Import" buttons in the engine manager are for? But there is also an "Export" button. Doesn't this already do what you want?

No. It exports only in the WBTM's native format which is not compatible with Winboard. I pester the gui geniuses until they include a Winboard inport feature for those of us who have extensive winboard ini files and just could not bear to enter the engines anew.

Ease of use cannot be overstated as a reason for not using a particular software - in this case Winboard.

For instance, I found it difficult to think about using Matthias Gemuh's rapidly improved (improving) TM until he incorporated an import ini feature in it.

I admit that ease of use has always been your central thesis.



Later.

[Zach Wegner]

So if it is not open source, or indeed not GPL, I foresee legal problems.

Is SMIRF open source? I guess it's just smirfoglot you're planning on including then?

BTW it might be nice to have an additional package of a bunch "partner" engines that can be downloaded, with some initial configuration setup in advance. There are plenty of decent GPL engines out there... </shameless self promotion>

[Marc Lacrosse]

I would offer my medium book which is permanently updated (...)

The book is a definite yes (...)

May I remind you that my performance.bin polyglot opening book which was initially bundled with Fruit 2.2 in 2005 has been made freely available and is downloadable at Leo's site since early 2006.

This is the book with which Fruit topped the SSDF rating list and won WBEC ridderkerk tournament a few months before rybka appeared.

Although it is now old (three years is an eternity in opening theory), this book still holds his own against most general opening books.

Just to see, I tested it against the one that Guenther Simon just offered to include in your proposed bundle (GS_medium081104).

I made two identical teams and had them play a scheveningen-type tournament. Team one played with Guenther's book and Team Two with my old performance.bin. Each member from team one played 50 fast blitz games against each member of team two on a fast quad under WinXP 64.

Each team was made of the four best free engines according to latest CCRL lists :
Rybka (2.2n2), Toga (1.4.2JD), Glaurung (2.1), Bright (0.3a).

Results :
Performance beats GSmedium +272 =276 -252 (51.25%)

This is almost equality according to bayeselo (+ 8 Elo points for Performance, with a LoS of 77%).

Taking into account that GSmedium weights 3271 KB while Performance is 1453 KB, I am not sure that your choice is best.

Regards

Marc

[H.G.Muller]

So if it is not open source, or indeed not GPL, I foresee legal problems.

Is SMIRF open source? I guess it's just smirfoglot you're planning on including then?

Indeed, only the adapter. I admit it is questionable if it is useful to include itin this release. OTOH, it is nearly zero size, and people that do want to play variants are usually not aware that it exists, although they are all aware of SMIRF. And it is just a few KB.

BTW it might be nice to have an additional package of a bunch "partner" engines that can be downloaded, with some initial configuration setup in advance. There are plenty of decent GPL engines out there... </shameless self promotion>

This might indeed be a good idea. But one thing at a time...

[Guenther Simon]

I would offer my medium book which is permanently updated (...)

The book is a definite yes (...)

May I remind you that my performance.bin polyglot opening book which was initially bundled with Fruit 2.2 in 2005 has been made freely available and is downloadable at Leo's site since early 2006.

This is the book with which Fruit topped the SSDF rating list and won WBEC ridderkerk tournament a few months before rybka appeared.

Although it is now old (three years is an eternity in opening theory), this book still holds his own against most general opening books.

Just to see, I tested it against the one that Guenther Simon just offered to include in your proposed bundle (GS_medium081104).

I made two identical teams and had them play a scheveningen-type tournament. Team one played with Guenther's book and Team Two with my old performance.bin. Each member from team one played 50 fast blitz games against each member of team two on a fast quad under WinXP 64.

Each team was made of the four best free engines according to latest CCRL lists :
Rybka (2.2n2), Toga (1.4.2JD), Glaurung (2.1), Bright (0.3a).

Results :
Performance beats GSmedium +272 =276 -252 (51.25%)

This is almost equality according to bayeselo (+ 8 Elo points for Performance, with a LoS of 77%).

Taking into account that GSmedium weights 3271 KB while Performance is 1453 KB, I am not sure that your choice is best.

Regards

Marc

Hi Marc,

Some thoughts here:

In my understanding HG seems to look for sth even smaller and
I created a 663KB(unzipped) Polyglot book file meanwhile, with
a max depth of 16 plies. But of course it is also just a well meant offer
and I don't care if any other book is included in that bundle.

I also want to mention that you can well get the same result you got
between my book and your book, if you just rename your book and let
it play against itself, because an exact 50% score in this test would
be even more unlikely statistically. I guess also that your book goes
further than max. 30 plies like mine? But this just as a side note.

Actually in the beginning I hated to create books, because IMHO it
should be the authors task to create a book for his engine for those
people who still like to match progs as they come, the oldfashioned
way with _own_ books. IMHO the author that way simply shifts the responsibility
for some non trivial part of the engines games to the poor users,
even if they didn't ask for it.

This happened more and more with UCI engines coming up(I am very
grateful therefore that Fruit the prototype of a great UCI engine came with
your book and that even the mother of all UCIs, Shredder, in later
versions delivered an own GUI independent book!
This is true of course for all other UCI engines authors who didn't decide against bothering
with integrating book code!) and now one can see lots of wasted games in the net because users
set their GUI books up wrongly or use faulty own book creations
or use settings not delivered by a third parties book creator or whatever...
AFAIK CCRL and CEGT delete all games with a too bad opening lines,
as they are responsible for the choosen book after they don't use
authors books at all. This is at least consequent. I hate that I have to do
the same when a bad line slipped through which happens very rarely,
but well as I said now the testers are responsible, which is not so nice.
Some testers simply ignore this, choose a book from a third party, may
be even set it up faulty and let the result even after -3.00 or bigger for
the first out of book move...
The author of such an engine probably just says never
mind the user provided crappy opening lines...

But well the situation cannot be changed anymore and now we have a lot
of programs(especially UCI) which need books from another side.
This creates a new problem at least for 'oldfashioned' (natural way?)testers,
who don't like to test from start positions or mini default books but
own 'normal' books. No single person can provide high quality _specialized_ books
for all of those bookless UCI engines. Instead the
compromise is to make a book which can _both_ hold its own against
books of engines with released books _and_ against _itself_ which is
nearly a paradoxon. This is because now all bookless UCI engines play
against the same used book...)
IMHO the way to test such a book against another book is to play it against
itself and measure the bad lines coming out of a sample of games
and playing both books against different other books and then compare
the results. I guess in the first part of the test a book with bigger depth
will have worse results and in the second part it's the opposite.
I hope I could make my thoughts and the dilemma somehow clear?
[not so sure about it ]

Regards,
Guenther

P.S. Despite all above I would still welcome a way to set max depth of
a Polyglot book via the Polyglot.ini file too, quasi simulating
book depth limiting by other UCI able GUIs.

Edited:
P.S.2 Oops, I forgot to ask for the played games against my medium book.
Of course I am interested in the lines played for further improving.
Thus if you still have the games available and want to share them,
I would be very pleased.
You could send them to rwbc(at)gmx.de

[Marc Lacrosse]

Dear Guenther

I like to distinguish three different kinds of opening books according to the main goal that one tries to reach while building and tuning a book:

Class A books : Here the goal is to offer to the amateur an opening book that covers a large variety of openings according to common opening theory so that a variety of games can be played on a solid ground : the engine will not be left on its own too early in any opening, no major opening is excluded but no really dubious line is included. Such a book will be short in minor openings (Bird opening) and will go far in major ones (spanish, najdorf ...). Good examples of such books are those that are produced by Chessbase and do come with their engines. This is also what you get if you build a PG book from a good selection of master games letting PG exceptional heuristics do its job.

Class B books : these are deliberately shorter books that are built for engines testing. The goal is to go into any significant "chess landscape" and let engines fight as soon as possible. Using these books is similar to testing from fixed positions "à-la-Noomen" but with a larger variety of openings. Short books by H Schnapp and S Canbaz belong to this category of books where a maximal number of moves is fixed (generally anywhere between seven to twelve moves).

Class C books: A book full of traps, counter-traps and secret preparation with a very narrow selection of overanalysed hand-tuned openings. These books are prepared for a single engine. The goal is to maximise this engine's result against more or less known opponents. In such a book, not only is each endpoint supposed to be a "good position" (in an absolute way) but the endpoints should be positions where you hope that this precise engine will play better chess than its opponents. Chosen lines are even better if they include possible traps for your opponents.

Then you may add a strange bird :

Class D books: similar to class C, but supposed to work for any engine against any opponent at any speed of play on any machine. The goal being to reach the highest percentage of success. To this category belong all those playchess-born books who fight for being best in achieving highest ranking at online blitz play ( sedat's "Perfect" books, and many many others).

Any book belongs to one of these categories (although some have a little touch of a second category).

For example, performance.bin was deliberately built as 80% class "A" and 20% class "C". Mainly class A because it was designed to be distributed for the pure amateur wishing to play and analyse with Fruit.
20% class C because I did a little tuning to favor openings with which Fruit had better results against main known computer opponents. This was asked by Fabien because the SSDF would test Fruit with this book.

What could we put in Winboard's bundle ?

I agree with you on the point that class D (best-performing for anyone against anyone) is essentially non-sensical.

I do not see anything interesting in distributing a class C book (for which engine? against which opponent?). A universal winning book does not exist.

Both yours and mine seem to be class A books. I just wished to stress that although yours is larger and more recent it does not seem to be "more correct" or "more strong" than my older freely available one.

We could also offer a class B one (a shortie for neutral engine tournaments). If you wish I commonly use myself a 20-plies one that I built myself. It can be distributed.

We could even distibute an EPD suite to be used as start-positions for engine matches and tournaments.
If you are interested my MLmfl suite is on freely available on the web as are those suites by J Noomen, A Silver and many others.

Best regards

Marc

PS Your idea of getting the opportunity to set max depth of a PG book through the ini file has limited interest IMHO because :
- You may do it when you build the book
- I am not sure that putting an endpoint to all lines at the exact same number of plies is the best way to get an interesting book.

[Guenther Simon]

Dear Guenther

I like to distinguish three different kinds of opening books according to the main goal that one tries to reach while building and tuning a book:

Class A books : Here the goal is to offer to the amateur an opening book that covers a large variety of openings according to common opening theory so that a variety of games can be played on a solid ground : the engine will not be left on its own too early in any opening, no major opening is excluded but no really dubious line is included. Such a book will be short in minor openings (Bird opening) and will go far in major ones (spanish, najdorf ...). Good examples of such books are those that are produced by Chessbase and do come with their engines. This is also what you get if you build a PG book from a good selection of master games letting PG exceptional heuristics do its job.

Class B books : these are deliberately shorter books that are built for engines testing. The goal is to go into any significant "chess landscape" and let engines fight as soon as possible. Using these books is similar to testing from fixed positions "à-la-Noomen" but with a larger variety of openings. Short books by H Schnapp and S Canbaz belong to this category of books where a maximal number of moves is fixed (generally anywhere between seven to twelve moves).

Class C books: A book full of traps, counter-traps and secret preparation with a very narrow selection of overanalysed hand-tuned openings. These books are prepared for a single engine. The goal is to maximise this engine's result against more or less known opponents. In such a book, not only is each endpoint supposed to be a "good position" (in an absolute way) but the endpoints should be positions where you hope that this precise engine will play better chess than its opponents. Chosen lines are even better if they include possible traps for your opponents.

Then you may add a strange bird :

Class D books: similar to class C, but supposed to work for any engine against any opponent at any speed of play on any machine. The goal being to reach the highest percentage of success. To this category belong all those playchess-born books who fight for being best in achieving highest ranking at online blitz play ( sedat's "Perfect" books, and many many others).

Any book belongs to one of these categories (although some have a little touch of a second category).

For example, performance.bin was deliberately built as 80% class "A" and 20% class "C". Mainly class A because it was designed to be distributed for the pure amateur wishing to play and analyse with Fruit.
20% class C because I did a little tuning to favor openings with which Fruit had better results against main known computer opponents. This was asked by Fabien because the SSDF would test Fruit with this book.

What could we put in Winboard's bundle ?

I agree with you on the point that class D (best-performing for anyone against anyone) is essentially non-sensical.

I do not see anything interesting in distributing a class C book (for which engine? against which opponent?). A universal winning book does not exist.

Both yours and mine seem to be class A books. I just wished to stress that although yours is larger and more recent it does not seem to be "more correct" or "more strong" than my older freely available one.

Hi Marc, I agree with most above. A note though my book is just
larger in size, because of a bigger base of newer games I guess,
but it is not as deep, as yours as I deliberately fixed it to max 30 plies
for Black and 28 for White.

We could also offer a class B one (a shortie for neutral engine tournaments). If you wish I commonly use myself a 20-plies one that I built myself. It can be distributed.

Well, I am not in the position to wish anything and if you want
your book to be distributed you just need to ask HG or what the
majority of the community thinks about a proper size/depth for
such a default book.

We could even distibute an EPD suite to be used as start-positions for engine matches and tournaments.
If you are interested my MLmfl suite is on freely available on the web as are those suites by J Noomen, A Silver and many others.

Yes, I think that is a good idea too, I even think adding an extra tool
like Odds GradualTest (for UCI and WB) for processing complete epd files
wouldn't be a bad idea either. I use it from time to time with one click
on a created batch file.

Something like this(which analysed collected positions from my last
team match for all our players):

Code: Select all

GradualTest /i RT1_2.epd /o Rybka_3-32.epd /t 600 /u /s "setoption name Hash value 512" Rybka_3-32/Rybka_3-32.exe >RT1_2.txt

Best regards

Marc

PS Your idea of getting the opportunity to set max depth of a PG book through the ini file has limited interest IMHO because :
- You may do it when you build the book
- I am not sure that putting an endpoint to all lines at the exact same number of plies is the best way to get an interesting book.

I think it is of interest for normal users, who don't create the Polyglot
book themselves, they could shrink a bigger book to any depth they like.
But that's more a generic approach...not what I would do myself except
for new experiments.

BTW because I edited my previous post very late, you probably missed
the point about my plea for the games of your test

Best regards,
Guenther

[Marc Lacrosse]

BTW because I edited my previous post very late, you probably missed
the point about my plea for the games of your test

Done


Marc

[Marc Lacrosse]

(...) if you want your book to be distributed (...)
[

Oh no.

It is already part of SCID distribution, Fruit distribution and is available at Leo's, at superchessengines and elsewhere.
So i do not want anything.
Simply this book still seems to hold a little value and is freely available.

Marc