Cutechess-cli 0.4.0 released

[IanO]

I've not used qmake before. Which project contains it? (Trying to make a Mac build.)

[ilari]

I've not used qmake before. Which project contains it? (Trying to make a Mac build.)

Qmake is one of the tools included in Nokia Qt SDK, which is needed to compile Cute Chess: http://qt.nokia.com/downloads/
The SDK is pretty big because of all the mobility stuff, Symbian and Meego simulators, etc.

[ilari]

Okay, cutechess-cli 0.4.1 is now available for Windows users here: http://koti.mbnet.fi/~ilaripih/bin/cute ... -win32.zip
This should fix the regression reported by Ferdy (big thanks!) that caused problems with reading engine output in Windows. Linux users can keep using v0.4.0.

Ferdinand, please let me know how things go with the new version. I'd also suggest using a very small time increment or time margin (it's a new option) with the N2 engine because it sends its move sometimes about 1 millisecond late.


Also... Robert Houdart, Ben-Hur Carlos Langoni, Lucas Braesch - you're welcome
And if you guys ever want to save whatever is left of your mental health, I suggest forgetting about writing engines and switching to chess (G)UI development - way more rewarding!

[Ferdy]

Okay, cutechess-cli 0.4.1 is now available for Windows users here: http://koti.mbnet.fi/~ilaripih/bin/cute ... -win32.zip
This should fix the regression reported by Ferdy (big thanks!) that caused problems with reading engine output in Windows. Linux users can keep using v0.4.0.

Ferdinand, please let me know how things go with the new version. I'd also suggest using a very small time increment or time margin (it's a new option) with the N2 engine because it sends its move sometimes about 1 millisecond late.


Also... Robert Houdart, Ben-Hur Carlos Langoni, Lucas Braesch - you're welcome
And if you guys ever want to save whatever is left of your mental health, I suggest forgetting about writing engines and switching to chess (G)UI development - way more rewarding!

Thanks, works fine now , no time loses on Deuterium vs Daydreamer and Deuterium vs Crafty both on TC 40moves/10sec.

There are still games decided due to time loses vs N2 that is without using timemargin option. Only time loses for N2 and not for Deuterium. I will just drop N2 as sparring partner done on TC below 40moves/1minute.

[ilari]

Thanks, works fine now , no time loses on Deuterium vs Daydreamer and Deuterium vs Crafty both on TC 40moves/10sec.

Glad to hear that. The bug was actually quite tricky to fix, like multithreading-related stuff often is.

There are still games decided due to time loses vs N2 that is without using timemargin option. Only time loses for N2 and not for Deuterium. I will just drop N2 as sparring partner done on TC below 40moves/1minute.

I used tc=40/10+0.02, and N2 didn't lose on time anymore, except for maybe once.

[ilari]

Surely I do something wrong but with a Polyglot book all openings are identical. Here is my batch file:

Code: Select all

cutechess-cli -fcp cmd=Houdini_15a.exe dir=e:\Sakk\UCI-engines\Houdini proto=uci option.Threads=1 -scp cmd=Vitruvius_082b.exe dir=e:\Sakk\UCI-engines\Vitruvius proto=uci -both tc=40/0:2 book=e:\Sakk\WinBoard-TM\WinBoard\Perfect2010.bin -draw 100 50 -resign 6 1000 -pgnout HouVit.pgn -games 100

Can anyone help me?

Hmm, it seems it does not happen using another book.

Perhaps the Perfect2010 book has weighted certain lines so heavily that cutechess-cli has no choice but to play them almost every time.

[ilari]

I have just checked a tournament I played under WinBoard using the same (Perfect2010) Polyglot book. It seems there were no repeated games.

Yeah, I just noticed this too. This problem is also somehow limited to the Windows version. Very strange, considering we don't have almost any platform-specific code outside of the inter-process communication classes. I'll investigate further...

[ilari]

I have just checked a tournament I played under WinBoard using the same (Perfect2010) Polyglot book. It seems there were no repeated games.

Yeah, I just noticed this too. This problem is also somehow limited to the Windows version. Very strange, considering we don't have almost any platform-specific code outside of the inter-process communication classes. I'll investigate further...

Found it: the problem is the ridiculously small range of random numbers produced by the rand() function in MSVC's libraries. Looks like I'll be ditching rand() completely. I hope I'll be able to release an update soon.

[ilari]

Found it: the problem is the ridiculously small range of random numbers produced by the rand() function in MSVC's libraries. Looks like I'll be ditching rand() completely. I hope I'll be able to release an update soon.

Cutechess-cli 0.4.2 released. The opening book move selector should now work correctly on all platforms. It used to work in Windows too, before I switched my compiling environment from GCC to MSVC. Thank you Gabor for reporting the bug. I updated the Linux binaries too, just in case.

Linux 32-bit: http://koti.mbnet.fi/~ilaripih/bin/cute ... x32.tar.gz
Linux 64-bit: http://koti.mbnet.fi/~ilaripih/bin/cute ... x64.tar.gz
Windows 32-bit: http://koti.mbnet.fi/~ilaripih/bin/cute ... -win32.zip