Sometimes when I analyze a position with my Old Rybka and it takes some longer to find the right move (I accept "right move" is a 'relative' term but still...)...then I switch off analysis and restart analysis...and gain switch off and restart...
When I do this funny procedure several times ... It now picks the Right Move or say best move it would have found out after five minutes of search...
what is mystery behind this???
regards
priyanvada
When I do this funny procedure several times ... It now picks the Right Move or say best move it would have found out after five minutes of search...
what is mystery behind this???
regards
priyanvada
You probably didn't unload the engine.
Yes I didn't...
But how does it help to find it the right move much quickly???
But how does it help to find it the right move much quickly???
Hash.
I c ...if this is the case then...
If we make adjustment in program such that it switches its analysis on and off after 3-5 seconds then it would perform much better than the current engine in our hand!
What say?!
Priyanvada
If we make adjustment in program such that it switches its analysis on and off after 3-5 seconds then it would perform much better than the current engine in our hand!
What say?!
Priyanvada
> If we make adjustment in program such that it switches its analysis on and off after 3-5 seconds then it would perform much better than the current engine in our hand!
No.
Is it just emotional NO or rational NO?!!
warm regards,
priyanvada
warm regards,
priyanvada
> Is it just emotional NO or rational NO?!!
LOL, u better ask, you won't know these things until you find out :)
;)
A no is a no. Stopping the search loses some information. Losing that information doesn't help.
As per my experience with such experiment...
20% if times losing that information doesn't help...
and 80% of the time it helps to ignore it!!!!
So I will still persist...
Someone master in chess programming needs to experiment on this...
Take normal engine and make an ON-OFF engine from the same CODE...
And play some sample matches between them, and let the Truth come out!
What say...!
regards
Priyanvada
20% if times losing that information doesn't help...
and 80% of the time it helps to ignore it!!!!
So I will still persist...
Someone master in chess programming needs to experiment on this...
Take normal engine and make an ON-OFF engine from the same CODE...
And play some sample matches between them, and let the Truth come out!
What say...!
regards
Priyanvada
> A no is a no. Stopping the search loses some information. Losing that information doesn't help.
Well you lose some information, yes, I suppose that could be true? That is not always a bad thing. What Priyanvada is doing is akin to Internal Iterative Deepening (IID). I would call it maybe MIID, Manual Incremental Iterative Deepening. Not even Vas knows about this yet, but it is going to be the next big thing for Rybka 4. There is going to be a new node-counter standard with a rate of one node per Rybka growthspurt. With an animation of Rybka slightly growing, splashing and then jumping out out of its fishbowl when it becomes too small, the user is required to restart the search by hitting the big blue 'Fresh Search'- button. If all goes well Rybka's old fishbowl in the upper left corner is replaced with a new, slightly bigger one, just before she dives in again for the next deep iterative search. The different searches can also be used for search sampling, Monte Carlo analysis and multivariational progress analysis of the evals to determine draw probabilities. This is a very powerful tool really!
Eelco
Thanks Eelco...
at least there is one who understands me!
Anyway whatever you have written has gone exactly parallel to my head...
Someday I will understand you also!
Someday...
regards
priyanvada
at least there is one who understands me!
Anyway whatever you have written has gone exactly parallel to my head...
Someday I will understand you also!
Someday...
regards
priyanvada
The hashtables will be constantly filled up with more and more search infos between the runs,
as long as you don't close the engine completely. This means it 'remembers' what it has already
found before and reuses this 'knowledge'.
It would find something even faster, if you won't stop analysis and restart it again, but go forward
and backward in your analysis by using already found good moves.
Guenther
as long as you don't close the engine completely. This means it 'remembers' what it has already
found before and reuses this 'knowledge'.
It would find something even faster, if you won't stop analysis and restart it again, but go forward
and backward in your analysis by using already found good moves.
Guenther
thanks
Please read the post above sir...
What do you think about that?
regards
Priyanvada
Please read the post above sir...
What do you think about that?
regards
Priyanvada
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill