hashhit position analysis for improving evaluation

Discussion of chess software programming and technical issues.

Moderator: Ras

Han Chengye
Posts: 23
Joined: Sun Feb 08, 2009 3:54 am

hashhit position analysis for improving evaluation

Post by Han Chengye »

collecting the high percent hashhit position, and analyze it, may help improve evaluation or help search?

anyone think about it?

sorry for bad english :oops:
Dann Corbit
Posts: 12792
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: hashhit position analysis for improving evaluation

Post by Dann Corbit »

Han Chengye wrote:collecting the high percent hashhit position, and analyze it, may help improve evaluation or help search?

anyone think about it?

sorry for bad english :oops:
A hash hit is often used for many valuable purposes.

Null move search sometimes examines the hash score to alter the depth of search.

If the hash hit is deep enough, no search is performed.

If the hash hit is not deep enough, it can be used to decide which node to search first.

What is the exact use that you were considering?
Han Chengye
Posts: 23
Joined: Sun Feb 08, 2009 3:54 am

Re: hashhit position analysis for improving evaluation

Post by Han Chengye »

what i mean is there're many many positions stored in hashtables which are generated by search, but those will be gone.
we can log it for later analyzing, to improving the engine's evaluation.
i don't know whether i say it clear or not:oops:
Dann Corbit
Posts: 12792
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: hashhit position analysis for improving evaluation

Post by Dann Corbit »

Han Chengye wrote:what i mean is there're many many positions stored in hashtables which are generated by search, but those will be gone.
we can log it for later analyzing, to improving the engine's evaluation.
i don't know whether i say it clear or not:oops:
So you mean to store the hash table to disk and reanalyze it later to learn from it.

I don't think anyone is doing that.