UCI question - limiting search depth

[royb]

I've successfully run some xboard matches with varying search depths (fixed depth) for Crafty-23.0 using the "sd=n" parameter in the .craftyrc file.

Now I would like to run a similar set of matches for some UCI engines (again, under xboard and using polyglot to handle the UCI to xboard conversion). I can get the UCI engine to run under xboard just fine, but it always seems to search to whatever depth it sees fit based on the time control. I'm not sure of how to tell the UCI engine how to limit its search depth a a fixed number of ply. I am using polyglot and the attempting to set "depth = 8" in the polyglot.ini file but that is not working. I see the UCI spec mentions "go depth 8" as the way to interactively force an 8 ply search. How do I control that for every move via some setting in the polyglot.ini file?

Also not working is using "initString ..." and secondInitString ... for the xboard command that starts the match when using UCI engines and polyglot.

Can someone point me in the right direction? The initString method should work for non-UCI engines -- it's the UCI portion that is throwing me I think.

Thanks,

Roy

[IWB]

Hello

I've successfully run some xboard matches with varying search depths (fixed depth) for Crafty-23.0 using the "sd=n" parameter in the .craftyrc file.
Now I would like to run a similar set of matches for some UCI engines (again, under xboard and using polyglot to handle the UCI to xboard conversion)...

I have no Idea about XBoard and Polyglott, but the UCI function itself works flawless in a UCI-GUI (Shredder-Classic) as long as this is really supported by the engine.

In your case there are 2 possibilities:

1. the Enigne you are trying to run to a certain depth is not supporting this
2. Ployglott has a problem - somehow

What are you trying to achieve with that - or better what you cant to is comparing the results for depth x for two engines. There are popular engines on the market who are hinding their true depth - or to make it more neutral: The "how to count depth method" can vary from engine to engine.

Bye
Ingo

[hgm]

The Polyglot source I have (Polyglot 1.4.38b) seems to handle the depth limiting in the proper way: it reacts to the sd N command (which is sent automatically when you run WinBoard with the -depth N or -searchDepth N option, but can also be included in the InitStrings to create depth-odds matches), and includes depth N in the UCI go command after having received it.

I don't know which Polyglot you are using.

[royb]

The Polyglot source I have (Polyglot 1.4.38b) seems to handle the depth limiting in the proper way: it reacts to the sd N command (which is sent automatically when you run WinBoard with the -depth N or -searchDepth N option, but can also be included in the InitStrings to create depth-odds matches), and includes depth N in the UCI go command after having received it.

I don't know which Polyglot you are using.

I am using Polyglot 1.4-3 under Ubuntu Linux. I tried using the initString and secondInitString method but Fruit-2.3.1 happily calculates to 13 or 14 ply. I wanted to make sure this worked right so I set depth to 1 ply which should make for instantaneous moves. My intent is to estimate how much each ply increase increases playing strength so I want to match ply 1 Fruit against ply 2 Fruit and then bump it up to around 12 ply in the end.

Here is my xboard command line:

xboard -initString "new\nsd 2\nrandom\n" -secondInitString "new\nsd 1\nrandom\n" -pixmap /home/roy/xboard/meridab/cocoa -xclock -xponder -mg 500 -size bulky -sgf /home/roy/chess/fruit-2ply-vs-fruit-1ply.pgn -fcp /home/roy/chess/fruit-2.3.1/linux/polyglot -fd /home/roy/chess/fruit-2.3.1/linux -scp /home/roy/chess/fruit-2.3.1/linux/polyglot -sd /home/roy/chess/fruit-2.3.1/linux

I have also tried creating two separate polyglot.ini files -- that also did not seem to work.

Thanks,

Roy

[hgm]

Well, I haven't seen the code of polyglot 1.4-3, but it is ancient.

It looks like you have a Polyglot problem, though, as the xboard command seems OK. So it would probbly be wise to first upgrade to Polyglot 1.4.39b, to see if the problem persists. And if it does, produce a Polyglot log file and post it hee, to see what is sent to the engine, and how this is wrong so that it does not work. It could of course be that the engine does not implement the depth qualifier to the go command.

In no Polyglot I know this is supposed to work from the polyglot.ini.

[royb]

Well, I haven't seen the code of polyglot 1.4-3, but it is ancient.

It looks like you have a Polyglot problem, though, as the xboard command seems OK. So it would probbly be wise to first upgrade to Polyglot 1.4.39b, to see if the problem persists. And if it does, produce a Polyglot log file and post it hee, to see what is sent to the engine, and how this is wrong so that it does not work. It could of course be that the engine does not implement the depth qualifier to the go command.

In no Polyglot I know this is supposed to work from the polyglot.ini.

I'm wiling to try the latest polyglot, but my google searches for polyglot 1.4.39b do not lead me to a download link. I will need source for a Linux (Ubuntu) platform, hopefully that is available.

Roy

[hgm]

http://alpha.uhasselt.be/Research/Algeb ... t-release/

There are sources there, but also binary debian and rpm packages. We also give this link from the GNU XBoard homepage.

While you are upgrading, you might also consider upgrading to WinBoard 4.4.0. That should make it easier to set the UCI or Polyglot options from the XBoard menu.

[hgm]

There is a debian binary package for the current (beta2) version of xboard 4.4.0 in the repository at:

http://home.hccnet.nl/h.g.muller/xboard ... 2_i386.deb

For other architectures you would have to pull the latest source tree from the git repository, and compile that using ./autogen.sh, ./configure and make.

[royb]

http://alpha.uhasselt.be/Research/Algeb ... t-release/

There are sources there, but also binary debian and rpm packages. We also give this link from the GNU XBoard homepage.

While you are upgrading, you might also consider upgrading to WinBoard 4.4.0. That should make it easier to set the UCI or Polyglot options from the XBoard menu.

Well, the new polyglot does not change the behavior. I have not yet tried the new polyglot in conjunction with xboard 4.4. That's next up.

Nope -- the new xboard and the new polyglot together still have the same problem.

Roy

[royb]

I also notice that with xboard 4.4 I get this error now:

Error -1 loading XPM file "/home/roy/xboard/merida/cocoa/fll72.xpm"

I never got that before. Is there a change to the pixmap formats in the new version? If so I won't use the new version as the default colors are just too ugly.

Roy