Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Info on 6 ply project

Author: Bas Hamstra

Date: 01:51:38 07/05/01

Go up one level in this thread


On July 05, 2001 at 03:30:06, Uri Blass wrote:

>On July 04, 2001 at 20:20:41, Bas Hamstra wrote:
>
>>On July 04, 2001 at 01:00:38, Uri Blass wrote:
>>
>>>On July 03, 2001 at 18:17:36, Bas Hamstra wrote:
>>>
>>>>On July 03, 2001 at 17:18:25, Uri Blass wrote:
>>>>
>>>>>On July 03, 2001 at 16:59:02, Bas Hamstra wrote:
>>>>><snipped>
>>>>>>Personally I have a problem believing it is possible to maintain such a rating
>>>>>>while searching 4 ply.
>>>>>
>>>>>I have no problem to believe that it is possible to maintain such a rating while
>>>>>searching 4 plies.
>>>>>
>>>>>You only need to call the rest of the plies in the name: "extensions".
>>>>>
>>>>>Uri
>>>>
>>>>That is hardly possible. At first I would say it is impossible to search only 4
>>>>4 "quiet" positional moves and still survive. What about a subtil tactic made
>>>>out of many quitet moves? But then again, Genius does quite an impressive job
>>>>with only 6 ply positional and a huge tactical search on top of that. Ok,
>>>>possible in principal, but very difficult.
>>>
>>>I guess that you do not have chessmaster6000
>>>It shows very small depthes in the first number
>>>chessmaster6000(ss=10) is even worse and often needs some minutes to get into
>>>depth 2.
>>>
>>>Chessmaster6000 shows 2 numbers.
>>>I guess that the first number is the brute force and the second number is search
>>>with null move pruning.
>>>
>>>depth 1/11 means 1 ply brute force and 11 plies with null move
>>>pruning+extensions so the main line is usually more than 11 plies.
>>>
>>>If you call search with null move pruning extensions you have no problem to see
>>>that chessmaster(ss=10) is searching usually less than 4 plies in the middle
>>>game.
>>>
>>>It is only a question of definitions
>>>
>>>Uri
>>
>>If you call search with nullmove pruning extensions? Except this is nonsense
>
>Why nonsense? Is there a rule to say what extensions can include?

There is consensus about what an extension is. When, in the full width part of
the search, you search a move deeper than normal on certain grounds, it is
called an extension. You are extending the normal search depth for that move.

Therefore to say "I call everything above 4 ply extensions" is nonsense, if we
want to stick to the normal meaning of the word "extension". If we invent our
own definitions all the time, it becomes a little difficult to have a
meaningfuill conversation.

>Suppose someone is using another selective algorithm without null move.
>Is there a rule when to call it extensions and when to call it depth?

Yes. 99.99% of all chess programs work by searching to a fixed depth and doing a
qsearch after that. If you extend that fixed depth search it is called an
extension :-)

>, I
>>think you misinterprete those 1/11 numbers.
>>
>>Bas.
>
>What is the meaning of them?
>
>My guess is that it means:
>1)1 ply of brute force and no null move pruning.
>2)11 plies of search with null move pruning
>3)A lot of extensions so practically chessmaster has no problem to see thins
>like mate in 18 at depth 1/11.

I am not sure, but my guess is this. In principal up till 11 ply it sees
everything that a normal 11 ply nullmove search would see. Plus, for the first 6
plies it looks for forced lines and follows them very deep (SE=6). To do this
heavy extending all the time would cost too much search depth, therefore he
restricts it to the plies near the root, where it's most dangerous.

Bas.













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.