Author: Antonio Dieguez
Date: 17:42:51 03/25/02
Go up one level in this thread
On March 25, 2002 at 19:09:30, TEERAPONG TOVIRAT wrote:
>On March 25, 2002 at 18:56:22, Gian-Carlo Pascutto wrote:
>
>>On March 25, 2002 at 18:41:52, TEERAPONG TOVIRAT wrote:
>>
>>>
>>>Hi,
>>>
>>>I've a few questions.
>>>
>>>1. When we extend searching depth,which one should be
>>>stored as hashtable->depth depth or depth+extension?
>>>
>>>
>>>Search(alpha,beta,depth)
>>>{
>>>
>>>Probehash()
>>>if(check) extension+=...
>>>
>>>-Search(-beta,-alpha,depth+extension-1)
>>>
>>>Storehash() => here with depth or depth+extension ???
>>>
>>>}
>>
>>You want things to be consistent. Think about this, is
>>should be easy to figure out for yourself.
>
>I guess u mean depth. Actually,I did a few tests.
>With depth it performs better. However,I don't understand
>the difference.
what a shame.
one example is if the original depth is x, you do an extension because you want
to search it with depth x+1. But if you store it with depth x+1 then at the next
iteration where it comes with depth x+1 and you want to search it to depth
x+1+1, the hash would be enough to make a hit or change bounds, wrongly.
I had the same mistake time ago. Not in my Search() but in my SearchFromCheck()
(i have them separated), and when I corrected it I also saw a small improvement
in test suites.
>>>2.After I add Extended Futility Pruning,cutting off at prefrontier
>>>nodes, to my program,it miss mate in 1.
>>>How to fix this?
>>
>>When you prune away moves, you have no more exact scores. You
>>need to store lower bounds into the hashtable.
>
>Unfortunately,I did that too.
I don't understand how can you miss a mate in 1 because a prunning.
anyway, you can store exact in the hash table, for ex. if you trust the moves
pruned would make the position below alpha, there is no problem.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.