Search

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
lauriet
Posts: 162
Joined: Sun Nov 03, 2013 8:32 am
Contact:

Search

Post by lauriet » Fri Jun 24, 2016 7:26 am

Hey Guys, (maybe Gals too....to be politically correct)

I have been trying to download/read various papers on search but would like the opinions of the experienced.....especially the academic (thats you Bob).
Of the various flavors of alpha/beta enhancements can anyone give any data on improvements of aspirational search and PVS.
eg: is aspirational 10% and PVS 20% faster ?

Thanks
Laurie

bob
Posts: 20549
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: Search

Post by bob » Fri Jun 24, 2016 5:58 pm

lauriet wrote:Hey Guys, (maybe Gals too....to be politically correct)

I have been trying to download/read various papers on search but would like the opinions of the experienced.....especially the academic (thats you Bob).
Of the various flavors of alpha/beta enhancements can anyone give any data on improvements of aspirational search and PVS.
eg: is aspirational 10% and PVS 20% faster ?

Thanks
Laurie
I could have answered that back in the 70's. But the answers are no good today since the search has changed so much. I suspect that PVS might still be worth 10%, but doubt if it goes very much higher. But as I said, that is based on 70's data.

The "enhancements" that seem most useful start with null-move, and include LMR and LMP, in addition to futility pruning... But I'd bet that if 10 of us disabled these one at a time, we'd get 10 different answers as to how much it impacted search.

Probably the best info can be found in "Scalable search in computer chess" written by Heinz. But he was not using all the things we use today so that will be a bit dated...

Post Reply