TalkChess.com

Hi all,

I have a new development version of Glaurung which I would like to see tested. Is there anyone who is willing to help?

I would be particularly interested in finding someone who also has some basic programming knowledge. The main addition compared to Glaurung 2.0.1 is king safety evaluation. The king safety evaluation is quite complex, and there's an awful lot of parameters to tune, far too many to make them all available as UCI parameters. It would therefore be nice to find someone who can modify the source code (I'll give advice about what to change, of course), compile, and run tests and experiments.

Of course, non-programmer testers would also be very welcome. Just knowing whether this version is stronger than 2.0.1 would be very interesting.

At the moment, I only have binaries for Linux and Mac OS X, but I expect to have 32-bit and 64-bit Windows binaries some time soon.

If anyone is interested, please drop me an e-mail, or send me a personal message on this board.

Tord

I'd be happy to help, but have you considered externalizing these parameters by making them settable through a config file? That may slow down runtime access a bit (since the settings will be variables not constants) but it makes it much easier to run tests with different combinations. I have done this recently with Arasan (it already had some configurable parameters, but now has more).

--Jon

always prepared to help you, but remember I am not programer neither I do one thousand test, I just play the ting and I give my subjective impressions.
If that is useful for you....
fernando

Shoot. From the thread topic I was hoping you needed testers for the iPhone version!

Tord Romstad wrote:Hi all,

I have a new development version of Glaurung which I would like to see tested. Is there anyone who is willing to help?

I would be particularly interested in finding someone who also has some basic programming knowledge. The main addition compared to Glaurung 2.0.1 is king safety evaluation. The king safety evaluation is quite complex, and there's an awful lot of parameters to tune, far too many to make them all available as UCI parameters. It would therefore be nice to find someone who can modify the source code (I'll give advice about what to change, of course), compile, and run tests and experiments.

Of course, non-programmer testers would also be very welcome. Just knowing whether this version is stronger than 2.0.1 would be very interesting.

At the moment, I only have binaries for Linux and Mac OS X, but I expect to have 32-bit and 64-bit Windows binaries some time soon.

If anyone is interested, please drop me an e-mail, or send me a personal message on this board.

Tord

Something else that would probably be helpful is a set of king safety test positions. That would at least show you that your changes have tactical value for those particular spots. Perhaps someone has collected such a set.

King safety links:
http://chess.about.com/library/weekly/aa03e17.htm
http://chess.about.com/od/beginners/ss/ble26pos_8.htm
http://query.nytimes.com/gst/fullpage.h ... A960948260
http://www.chesscorner.com/tutorial/Int ... ements.htm
http://www.associatedcontent.com/articl ... _king.html

Hi Dann,

Thanks for the links! I've only had time for a very quick look so far, but it is quite possible that I'll be able to find something useful when I study them more carefully.

Tord

Mark wrote:Shoot. From the thread topic I was hoping you needed testers for the iPhone version!

That wouldn't even be technically possible yet, because it still isn't possible to transfer applications made with Apple's iPhone SDK to an actual iPhone. The earliest possible date would be in the end of June (assuming that I have a complete and working program by then, which is far from sure), and at the moment it is not clear whether Apple will accept non-US based developers from the beginning.

At any rate, the iPhone version of Glaurung is still very much part of my plans, and I hope to have the first version ready well before the end of the year.

Tord

jdart wrote:I'd be happy to help, but have you considered externalizing these parameters by making them settable through a config file? That may slow down runtime access a bit (since the settings will be variables not constants) but it makes it much easier to run tests with different combinations. I have done this recently with Arasan (it already had some configurable parameters, but now has more).

A config file would definitely be worth implementing, but in this case it wouldn't be quite general enough for my needs. I was talking about "parameters" in a wide sense of the word: Not just constants, but also what formulas to use on the constants. Because C++ does not provide any way to compile and load code during runtime, I can't put actual code in a config file.

Tord

It is true you can't compile at runtime but you can always do:

if (option1) {
king safety code variant 1
}
else {
king safety code variant 2
}

I'm no programmer, but I'll be glad to test different parameters. I have a quad core, so I could test 4 different engine parameters at the same time through engine matches (ponder off, of course).