Each element in the countermove table is a history table in this version.
Dividing history after each move also worked:http://www.talkchess.com/forum/viewtopi ... 10&t=63870
But most changes were bugfixes in search and eval (e.g. eval is color symmetric now)
New strength should be above 2100 CCRL.
Galjoen 0.36
Moderators: hgm, Harvey Williamson, bob
Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
- Werner Taelemans
- Posts: 81
- Joined: Mon Feb 03, 2014 10:57 am
- Location: Belgium
- Contact:
- Graham Banks
- Posts: 30733
- Joined: Sun Feb 26, 2006 9:52 am
- Location: Auckland, NZ
Re: Galjoen 0.36
Is there an exe for download?
Graham.
Graham.
My email addresses:
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz
- Werner Taelemans
- Posts: 81
- Joined: Mon Feb 03, 2014 10:57 am
- Location: Belgium
- Contact:
Re: Galjoen 0.36
If you download the source, you will also find the map bin.
This map contains a 32-bit and 64-bit exe.
This map contains a 32-bit and 64-bit exe.
Re: Galjoen 0.36
I have used Galjoen 0.36 under Arena in both WB and UCI mode and it seems to me it plays weaker in WB mode.
Against a pool of opponents it scored 42 out of 102 in WB mode, then 83,5 out of 154 in UCI mode.
Against a pool of opponents it scored 42 out of 102 in WB mode, then 83,5 out of 154 in UCI mode.
Gabor Szots
CCRL testing group
CCRL testing group
- Werner Taelemans
- Posts: 81
- Joined: Mon Feb 03, 2014 10:57 am
- Location: Belgium
- Contact:
Re: Galjoen 0.36
Galjoen started as a CECP engine. Later I switched to UCI but kept support for CECP.
Today Galjoen is a UCI engine with an extra layer that takes care of the translation between CECP and UCI.
This extra translation seems to hurt more than I expected (especially at short time controls),
and a 13% difference is obviously unacceptable.
The easiest solution, for me of course, is to throw away support for CECP.
I could also try to optimize all this, but I'm afraid that for very short time controls there would still be a difference in strength.
My question: is there in fact interest or demand for engines that support both protocols?
Today Galjoen is a UCI engine with an extra layer that takes care of the translation between CECP and UCI.
This extra translation seems to hurt more than I expected (especially at short time controls),
and a 13% difference is obviously unacceptable.
The easiest solution, for me of course, is to throw away support for CECP.
I could also try to optimize all this, but I'm afraid that for very short time controls there would still be a difference in strength.
My question: is there in fact interest or demand for engines that support both protocols?
Re: Galjoen 0.36
I can't answer this question. I have never really understood the difference between these protocols. When an engine is UCI only, I use it under Shredder, Fritz, Arena and WinBoard alike. When it is WB only, I use it only under Arena and WinBoard. So it seems I have a bias for UCI but infact I only want to avoid external adapters.Werner Taelemans wrote: My question: is there in fact interest or demand for engines that support both protocols?
When an engine supports both protocols I use it as a WB engine under Arena for the very reason to detect implementation flaws. That's why I started my tournament with Galjoen WB.
Gabor Szots
CCRL testing group
CCRL testing group
- Werner Taelemans
- Posts: 81
- Joined: Mon Feb 03, 2014 10:57 am
- Location: Belgium
- Contact:
Re: Galjoen 0.36
Thanks for testing this. I put this on my to do list for version 0.37, because it's too complicated to do with a small patch now.SzG wrote:When an engine supports both protocols I use it as a WB engine under Arena for the very reason to detect implementation flaws. That's why I started my tournament with Galjoen WB.
