Author: Christophe Theron
Date: 17:24:21 05/15/00
Go up one level in this thread
On May 15, 2000 at 18:44:14, Ed Schröder wrote:
>On May 15, 2000 at 17:34:12, Bas Hamstra wrote:
>
>>Frans Morsch wouldn't really try alternatives for good old alphabeta, would he?
>>
>>By the way, I read somewhere that someone obtained good results doing alphabeta
>>near the root and best-first minimax near the leaves. Has anyone tried that, or
>>is it nonsense? Doing only best-first MM was apparently too selective.
>>
>>Regards,
>>Bas Hamstra.
>
>The authors of SSS (State Space Search) claim it is superior to
>alpha/beta. Maybe in the meantime there are improvements, who knows.
>I will have to read the article again.
>
>Ed
By giving infinite hash table, SSS* becomes equivalent to MTD(f) (they search
the same tree). This is an incredible result, but has been demonstrated by Aske
Plaat. SSS* requires to store all the visited nodes during the search, which
makes it practically unusable for real life. MTD(f) just requires to have a big
hash table, which is practically doable.
SSS* is not alphabeta, but MTD(f) is. So there is no need to give up alphabeta
in order to achieve the good results of SSS*.
Several good programs already use MTD(f) (I'm sure Anmon does, and I remember
that Don Dailey was working hard on it for Clikchess one year ago).
MTD(f) or SSS* is more effective than PVS/Negascout, but it has some drawbacks.
Some pruning schemes become really unstable with MTD(f), as Christian Barreteau
and Don told us.
Maybe if Christian or Don reads this they could tell us more about their
experience with MTD(f)...
Christophe
This page took 0.01 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.