Author: Christophe Theron
Date: 01:02:41 08/17/00
Go up one level in this thread
On August 17, 2000 at 01:06:11, Peter Kappler wrote:
>On August 16, 2000 at 23:09:02, Christophe Theron wrote:
>
>>On August 16, 2000 at 19:47:27, Dan Andersson wrote:
>>
>>>On August 16, 2000 at 19:34:21, Gian-Carlo Pascutto wrote:
>>>
>>>>On August 16, 2000 at 18:52:03, David Rasmussen wrote:
>>>>
>>>>>My Qsearch usually takes up 70-80 % of the nodes searched.
>>>>>What are the, say, 10 best ideas/techniques to cut this number down?
>>>>
>>>>a) don't qsearch
>>>>
>>>>b) use futility pruning
>>>>
>>>>c) use SEE futility pruning
>>>>
>>>>d) live with it...qsearch % tends to be high in every chessprogram
>>>>
>>>>>P.S. I am writing a new chess program from scratch and I'm very much in love
>>>>>with the scientific beauty of MTD(f). What are the pratical pros and cons of
>>>>>doing MTD(f). What are the pitfalls etc. ?
>>>>
>>>>It causes trouble with search trics that depend on alpha/beta values.
>>>But introduces new tricks. Especially if one uses ETC.
>>
>>
>>Sorry to ask (actually you certainly expect someone will ask), but
>>
>>What is ETC?
>>
>>
>> Christophe
>>
>
>
>Enhanced Transposition Cutoffs.
>
>If the current position isn't in the hashtable, you try each legal move in the
>hope that one of them will produce a hash hit with a score that can cause a
>cutoff.
>
>I think Schaeffer(?) wrote a paper on this...
>
>--Peter
Thanks.
Christophe
>>>>Its harder to get a PV.
>>>Yep.
>>>>It's not necessarily faster than PVS.
>>>In the best case MTD and PVS will search the same nodes. And if you use ETC I'm
>>>almost certain it will search a smaller tree in the average case. ETC will slow
>>>search down, but there are ways to alleviate that. F.ex no ETC the last N plies.
>>>>Depends more on large hashtables than other methods.
>>>Í would say that it suffers badly from to small TTs instead, maybe even
>>>benefitting more from large TTs than other methods. (Thats pure speculation)
>>>>
>>>>>P.P.S. Why is my chess program code so messy?
>>>>
>>>>Because its a chessprogram.
>>>>
>>>>--
>>>>GCP
>>>
>>>Regards Dan Andersson
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.