Author: Koundinya Veluri
Date: 15:55:03 09/01/01
Go up one level in this thread
On September 01, 2001 at 17:23:02, Antonio Dieguez wrote:
>On September 01, 2001 at 17:19:44, Antonio Dieguez wrote:
>
>>
>>>int ev=evaluate(...)
>>>
>>>if (ev>=beta+window)
>>>{ return beta; }
>>>
>>>then see captures with this window: (alpha-window,beta)
>>>
>>>if that alpha-window was not improved then return alpha.
>>>
>>>(less nodes, and now exactly the same, if I am not mistaken...)
>>
>>it is not the same! but perhaps you like it better anyway.
>>
>>It is not the same because in your original code, when your quiesce return a
>>value *by evaluation*, inside the window, that pos. is not pruned then but now
>>it can... well, I give up! :)
>
>This way:
>
>int ev=evaluate(...)
>
>if (ev>=beta+window)
>{ return beta; }
>
>then see captures with this window: (alpha-window,beta)
>
>if that alpha-window was not improved AND ev<=alpha-window then return alpha.
>
>------------
>
>IT'S THE SAME. I sure! :) :)
Yes this is pretty much the same! Wow, I never thought it would be possible
without the extra qsearch, but you seemed to have conquered it!
Although it is missing just one type of pruning that my original code was doing,
the case where a capture causes alpha to increase beyond beta+window, but this
is a small thing.
I had some trouble understanding why the ev<=alpha-window was needed, but the
way I understood it, I think it would be better to use a wider window in the
evaluate() call itself rather than specify that ev<=alpha-window at the end.
Here is the modified stuff put together:
int ev=evaluate(alpha-window,beta+window)
if (ev>=beta+window)
{ return beta; }
then see captures with this window: (alpha-window,beta+window)
if that alpha-window was not improved then return alpha.
if that alpha-window was improved to greater than beta+window then return beta.
Think this should be right. I will let you know how "same" it really is after
testing. Thanks for your many suggestions!
Koundinya
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.