Author: Johan de Koning
Date: 14:27:17 03/27/04
Go up one level in this thread
On March 26, 2004 at 12:37:11, Robert Hyatt wrote:
>On March 25, 2004 at 22:52:02, Johan de Koning wrote:
>
>>On March 25, 2004 at 16:40:26, Robert Hyatt wrote:
>>
>>>On March 25, 2004 at 01:56:43, Johan de Koning wrote:
>>>
>>>>On March 24, 2004 at 11:09:41, Robert Hyatt wrote:
>>>>
>>>>>On March 23, 2004 at 05:05:56, Vasik Rajlich wrote:
>>>>
>>>>>>Junior, however, appears to come at the problem of selective search via
>>>>>>discussions about this in the CCC archives. Amir has claimed that the best way
>>>>>>to search selectively is via extensions. To complete the reductions vs
>>>>>>extensions thought from above, an extension strategy will have the profile that
>>>>>>most moves have the same basic search depth, while certain special moves will
>>>>>>have a higher search depth. The profile of a search based on reductions compared
>>>>>>to a search based on extensions will be different.
>>>>>
>>>>>It is easy to prove that last statement wrong.
>>>>>
>>>>>You write a program that only does search depth reductions. I write a program
>>>>>that only does extensions. I can make mine _identical_ to yours. Where you
>>>>>reduce, I do nothing. Where you don't reduce, I extend. IE if you don't reduce
>>>>>a check, I extend the check. We search _exactly_ the same tree.
>>>>
>>>>Indeed, assuming fractional plies, it is rather trivial to build
>>>>the same tree using either extensions or reductions.
>>>>
>>>>But it's better to avoid the term "reductions" since it is confusing.
>>>>The real issue is extensions versus *pruning*.
>>>
>>>Let me define _my_ vocabulary to avoid further confusion.
>>>
>>>1. Extension. extending the depth of a move based on some property it
>>>exhibits, such as being a check or whatever.
>>>
>>>2. Reduction. Reducing the depth of a move based on some property it exhibits,
>>>such as not being a capture, check, threat, etc.
>>>
>>>The two terms are inverses. I can extend the set of moves {X} or I can reduce
>>>the set of moves {M-X} and get _exactly_ the same result, to the node. Note
>>>that M is the set of all moves we will search.
>>>
>>>3. Forward-pruning. Taking some set of moves at the current ply and throwing
>>>them out with no additional searching of any kind.
>>>
>>>4. Backward-pruning. IE alpha/beta pruning that doesn't change the final
>>>result at all.
>>
>>Fair enough, but null moving doesn't fit in your vocabulary.
>
>Actually it does. It is a "reduction"... The reduction is "R" and it is done
>when the shallow search can't find bad after I "pass"...
It is if you consider null moves to be part of M, as I wrote 1 line down.
But the equivalence of extension based and reduction based searches will
start to look silly then.
>>One solution is to define null moves as part of the reference tree
>>(a search that utilizes 4. at most).
>>
>>Another way is to allow searches under 3. After all, null move is an
>>estimate *and* it is used to disqualify members of M. That's sounds
>>like pruning! :-) And after hiding the null searches in an (expensive)
>>black box there is no difference at all.
>
>Note it really doesn't prune, as in throwing things away with no search, it does
>a search to a reduced depth...
>
>Whether that is pruning or not is probably a religious vs technical argument...
>:)
Technically I'd say it is pruning unless you allow null moves in the PV.
And I think I'll skip the religious part for this once. :-)
... Johan
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.