Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: MTD(f) Problems

Author: Andrew Williams

Date: 01:25:24 04/12/02

Go up one level in this thread


On April 11, 2002 at 17:17:48, martin fierz wrote:

>On April 11, 2002 at 05:07:20, Andrew Williams wrote:
>
>>On April 11, 2002 at 03:58:01, martin fierz wrote:
>>
>>>On April 11, 2002 at 02:22:55, Alejandro Dubrovsky wrote:
>>>
>>>>On April 10, 2002 at 21:02:51, Dann Corbit wrote:
>>>>
>>>>>On April 10, 2002 at 19:40:45, Matthias Gemuh wrote:
>>>>>
>>>>>>Almost everyone here speaks AlphaBeta :(.
>>>>>
>>>>>Vincent uses MTD(f), IIRC.
>>>>>Smallpotato's author uses MTD(f).
>>>>>KnightCap uses MTD(f).
>>>>>
>>>>>There are maybe 6-7 others who like it.
>>>>
>>>>I use it, but i don't like it. Search instabilities seem harder to handle. Next
>>>>version of smallpotato will probably switch to aspiration search.
>>>>
>>>>Alejandro
>>>
>>>i use MTD in my checkers program. it searches smaller trees,  but also has
>>>drawbacks. i agree on the search instabilities, and AFAIK, you cannot get a PV
>>>with MTD - which makes understanding why it played a move harder... or is there
>>>a way to get a PV?
>>
>>I get mine out of my transposition table.
>>
>>Andrew
>
>so do i, but you have no guarantee that you get the whole PV, because an entry
>in the TT might have been overwritten...
>
>cheers
>  martin

Yeah. When I met Frans Morsch at a WMCC in London, he said he liked the idea of
MTD(f) but didn't like the idea of not being able to get the correct PV.

Andrew



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.