RedQueen v0.9.8 release

[bhlangonijr]

Another release before 1.0 beta.

I compiled your engine for Android, and put the binary on the UCI engines for Android page (v095 is also still available).

Thanks a lot, Aart!

[bhlangonijr]

Another release before 1.0 beta.

- SMP support up to 16 cores;
- Fixed getFEN method;
- Improved evaluation - Fixed some material imbalances;
- Fixed a bug at the SEE function
- Some refactoring at the Search class

Linux 32/64 bits binaries/source: http://sourceforge.net/projects/redquee ... p/download
Windows 32/64 bits binaries/source: http://sourceforge.net/projects/redquee ... p/download

SVN tag: https://redqueenchess.svn.sourceforge.n ... ease-0.9.8

Regards,

RedQueen's fast builds at: http://jimablett.hostwebs.com/

Thanks a lot Jim!

[michiguel]

Another release before 1.0 beta.

- SMP support up to 16 cores;
- Fixed getFEN method;
- Improved evaluation - Fixed some material imbalances;
- Fixed a bug at the SEE function
- Some refactoring at the Search class

Linux 32/64 bits binaries/source: http://sourceforge.net/projects/redquee ... p/download
Windows 32/64 bits binaries/source: http://sourceforge.net/projects/redquee ... p/download

SVN tag: https://redqueenchess.svn.sourceforge.n ... ease-0.9.8

Regards,

Looks like we are all getting ready for TCEC
Let's kick some "top engine's bottoms"

Miguel

Miguel, I tested (in the overnight) the new Gaviota against RQ and it seems it is a drastic improvement for Gaviota. I guess it is about 200 Elo stronger than 0.80.

Congrats!

Thanks BH, but it must have been an accident... 200 you say? I would settle for 100. Like we chatted in Chessbomb, there were some obvious changes in search that I needed to introduce and those certainly helped (only after tuning the parameters a lot!!).

Miguel

[abik]

I compiled your engine for Android, and put the binary on the UCI engines for Android page (v095 is also still available).

Thanks a lot, Aart!

My pleasure. Here are the result of an engine-engine match between the two versions with 1 second per move for 100 games (50 fixed book openings from both sides) 32MB hash size on a Nexus One phone.

Code: Select all

1   RedQueen 0.9.8  +37/-34/=29 51.50%   51.5/100
2   RedQueen 0.9.5  +34/-37/=29 48.50%   48.5/100

[bhlangonijr]

My pleasure. Here are the result of an engine-engine match between the two versions with 1 second per move for 100 games (50 fixed book openings from both sides) 32MB hash size on a Nexus One phone.

Code: Select all

1   RedQueen 0.9.8  +37/-34/=29 51.50%   51.5/100
2   RedQueen 0.9.5  +34/-37/=29 48.50%   48.5/100

The major changes between these two versions are actually related to the addition of SMP support. But there are changes on the evaluation function (added some minor rules to fix material imbalances) and also fixed a bug in the SEE function. Running self-matches using the single core version the changes are worth ~25 Elo.

Thanks!