(iteration+depth) TT replacement policy.

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Aleks Peshkov
Posts: 870
Joined: Sun Nov 19, 2006 8:16 pm
Location: Russia

(iteration+depth) TT replacement policy.

Post by Aleks Peshkov » Fri Feb 27, 2015 10:38 pm

Just caught an idea using (iteration+depth) as replacement metrics. The most important records should be visited in every iteration of iterative deepening, so the deepest entries would still have priority between most recent ones.

But outdated unused deep entries would be flushed out very fast.

Rein Halbersma
Posts: 685
Joined: Tue May 22, 2007 9:13 am

Re: (iteration+depth) TT replacement policy.

Post by Rein Halbersma » Sat Feb 28, 2015 10:58 am

How would that look in code?

mcostalba
Posts: 2684
Joined: Sat Jun 14, 2008 7:17 pm

Re: (iteration+depth) TT replacement policy.

Post by mcostalba » Sat Feb 28, 2015 12:06 pm

This may be an interesting idea to play with. Currently engines use aging (in case of SF is called 'generation' and is incremented every new search) to prefer entries from current search to older ones.

This would be a kind of tradeoff, not so strong as new generation, but still something to add to depth.

Aleks Peshkov
Posts: 870
Joined: Sun Nov 19, 2006 8:16 pm
Location: Russia

Re: (iteration+depth) TT replacement policy.

Post by Aleks Peshkov » Sat Feb 28, 2015 6:00 pm

Iteration counter grows very fast at the beginning of new generation, so unused entries would still be flushed out except the deepest ones.

Post Reply