Blog

Discussion of chess software programming and technical issues.

Moderators: hgm, Harvey Williamson, bob

Post Reply
User avatar
Rebel
Posts: 4088
Joined: Thu Aug 18, 2011 10:04 am

Blog

Post by Rebel » Thu Nov 16, 2017 4:37 pm

I opened a blog with ideas. Hope to update it on a frequently base.

http://rebel13.nl/rebel13/ideas.html

User avatar
Steve Maughan
Posts: 1024
Joined: Wed Mar 08, 2006 7:28 pm
Location: Florida, USA
Contact:

Re: Blog

Post by Steve Maughan » Thu Nov 16, 2017 6:22 pm

Bookmarked!

Thanks Ed
http://www.chessprogramming.net - Maverick Chess Engine

Gerd Isenberg
Posts: 2103
Joined: Wed Mar 08, 2006 7:47 pm
Location: Hattingen, Germany

Re: Blog

Post by Gerd Isenberg » Thu Nov 16, 2017 6:41 pm

Rebel wrote:I opened a blog with ideas. Hope to update it on a frequently base.

http://rebel13.nl/rebel13/ideas.html
Hi Ed,
is the new blog intended to replace the "Programmer Corner" or "Programmer Stuff" site, which "will disappear in the near future. " If yes, you may incorporate the old stuff inside the new blogs ...

http://www.top-5000.nl/authors/rebel/chess840.htm

I love it to update links in cpw ;-)

Cheers,
Gerd

phhnguyen
Posts: 254
Joined: Wed Apr 21, 2010 2:58 am
Contact:

Re: Blog

Post by phhnguyen » Thu Nov 16, 2017 9:17 pm

Rebel wrote:I opened a blog with ideas. Hope to update it on a frequently base.

http://rebel13.nl/rebel13/ideas.html
With a fan or hobby page in Facebook? (so we can be updated easily) :wink:

User avatar
Rebel
Posts: 4088
Joined: Thu Aug 18, 2011 10:04 am

Re: Blog

Post by Rebel » Thu Nov 16, 2017 10:42 pm

Gerd Isenberg wrote:
Rebel wrote:I opened a blog with ideas. Hope to update it on a frequently base.

http://rebel13.nl/rebel13/ideas.html
Hi Ed,
is the new blog intended to replace the "Programmer Corner" or "Programmer Stuff" site, which "will disappear in the near future. " If yes, you may incorporate the old stuff inside the new blogs ...

http://www.top-5000.nl/authors/rebel/chess840.htm

I love it to update links in cpw ;-)

Cheers,
Gerd
Hi Gerd,

I have added the old inside_rebel stuff in PDF to the new page.

User avatar
Rebel
Posts: 4088
Joined: Thu Aug 18, 2011 10:04 am

Re: Blog

Post by Rebel » Thu Nov 16, 2017 11:08 pm

phhnguyen wrote:
Rebel wrote:I opened a blog with ideas. Hope to update it on a frequently base.

http://rebel13.nl/rebel13/ideas.html
With a fan or hobby page in Facebook? (so we can be updated easily) :wink:
Why not :D

https://www.facebook.com/Rebel-415274595331860/

phhnguyen
Posts: 254
Joined: Wed Apr 21, 2010 2:58 am
Contact:

Re: Blog

Post by phhnguyen » Fri Nov 17, 2017 1:06 am

Rebel wrote:
phhnguyen wrote:
Rebel wrote:I opened a blog with ideas. Hope to update it on a frequently base.

http://rebel13.nl/rebel13/ideas.html
With a fan or hobby page in Facebook? (so we can be updated easily) :wink:
Why not :D

https://www.facebook.com/Rebel-415274595331860/
👍

Daniel Shawul
Posts: 3429
Joined: Tue Mar 14, 2006 10:34 am
Location: Ethiopia
Contact:

Re: Blog

Post by Daniel Shawul » Fri Nov 17, 2017 3:55 am

How much are the rybka material tables worth ?

If I add tuning to it, this method is equivalent to tuning a material configuration, instead of
tuning individual pieces. Maybe it is safer to tune for the perturbation of the total material with the base calculated from 9-5-3-3-1 score or something similar. This is good to overcome the problem when a material configuration does not occur in the database.

Daniel

User avatar
Rebel
Posts: 4088
Joined: Thu Aug 18, 2011 10:04 am

Re: Blog

Post by Rebel » Fri Nov 17, 2017 6:25 am

Daniel Shawul wrote: How much are the rybka material tables worth ?
No idea, yet. I am currently testing the safe-option (min 10 games with a 0% or 100% score using the number of games as bonus c.q. penalty with a maximum of one pawn. So far a small positive result.

Overall I think the gain won't be that much, maybe 10-15 elo max for mine provided I find the optimal formula. But you never know if you don't try.
Daniel Shawul wrote:If I add tuning to it, this method is equivalent to tuning a material configuration, instead of tuning individual pieces. Maybe it is safer to tune for the perturbation of the total material with the base calculated from 9-5-3-3-1 score or something similar. This is good to overcome the problem when a material configuration does not occur in the database.
Well of course the MIT is not meant to replace existing knowledge, it's an addition to fine tune the existing material configuration based on statistics. To overcome that missing patterns in the database may become a problem in certain positions use not too high bonus / penalties.

Using all the 5 million games gives a 50% fill which is close to having them all because the other half of the cells (or at least a great part) are about crazy positions with a queen and a handfull of pawns less which will never occur and if they do don't matter, no bonus / penalty needed.

Filling in the gaps with a 9-5-3-3-1 formula or something similar is tricky, I will keep it in mind for later.

User avatar
Rebel
Posts: 4088
Joined: Thu Aug 18, 2011 10:04 am

Re: Blog

Post by Rebel » Sun Dec 17, 2017 4:57 pm


Post Reply