Author: Christophe Theron
Date: 02:19:11 09/01/01
Go up one level in this thread
On September 01, 2001 at 02:56:44, Uri Blass wrote:
>On September 01, 2001 at 01:02:42, Christophe Theron wrote:
>
>>On August 31, 2001 at 15:57:28, Uri Blass wrote:
>>
>>>On August 31, 2001 at 13:05:35, Christophe Theron wrote:
>>>
>>>>On August 31, 2001 at 03:17:23, Peter Berger wrote:
>>>>
>>>>>On August 31, 2001 at 02:44:52, Christophe Theron wrote:
>>>>>
>>>>>>No I just assume better branching factor from my program.
>>>>>>
>>>>>>Tiger needs less time than Genius to complete each successive iteration, and the
>>>>>>effects of this better branching factor shows up more clearly when more
>>>>>>iterations (ply depths) are completed.
>>>>>>
>>>>>>In sports, we would say that Genius has more explosive power (is a good
>>>>>>sprinter) but gets tired very quickly. So if the race lasts longer, Tiger does
>>>>>>not get tired when Genius is exhausted already.
>>>>>>
>>>>>>You are not going to see this difference if you let them run only short races.
>>>>>>
>>>>>
>>>>>I don't disagree at all ( the diminishing return remark was supposed to be a
>>>>>joke) and everyone can see this effect on current CPUs. The question is at which
>>>>>speed this effect really kicks in .
>>>>>
>>>>>It it a 200 or a 400 meter race ? Or will ChessTiger need a marathon :) ?
>>>>>
>>>>>I expect ChessTiger will do well in sprints, too btw .
>>>>>
>>>>>Until the 486 came out Genius was probably still in active development. I assume
>>>>>_if_ it was optimized for _any_ speed it was probably tournament time control on
>>>>>a 386 or 486 CPU . Genius finds several very clever moves at fast times where
>>>>>modern programs probably will rely to get them anyway with search .
>>>>
>>>>
>>>>
>>>>It does not make any sense to talk about "optimizing for 386 or 486 tournament
>>>>time controls".
>>>>
>>>>It would take years to achieve a task like this, and while this optimization job
>>>>would take place the author would not be able to make any serious change in his
>>>>program.
>>>>
>>>>An author simply tries to make his program stronger, and that's already a task
>>>>difficult enough, from the human point of view.
>>>>
>>>>I do not know of any improvement that would be a blitz improvement only (I mean
>>>>an improvement that would only help in blitz and not at longer time controls).
>>>>Likewise, I do not know any improvement that would only help at long time
>>>>controls.
>>>>
>>>>"Optimizing for blitz or tournament time controls" is a fantasy in the mind of
>>>>many readers of this discussion group, but as many other things I see discussed
>>>>here, it does not exist.
>>>>
>>>>
>>>>
>>>> Christophe
>>>
>>>I am sure that there are changes that are productive at
>>>long time control but counter productive at short time control.
>>>
>>>For example simple logic says that a big book may be productive
>>>at short time control when a smaller book may be more productive
>>>at long time control.
>>>
>>>In the extreme case at infinite time control
>>>the engine can find all the right moves and
>>>needs no book
>>
>>
>>
>>Forget about the extreme cases, we will never get there.
>>
>>I don't see why a better book would be counter productive at longer time
>>controls. You should email your theory to Jeroen, Alex and Sandro.
>
>I am not against Jeroen,Alex and Sandro book but I believe that you can do your
>program stronger at blitz by adding a big blitz book when the program will play
>moves from the blitz book when it is out of the tournament book.
>
>The blitz book can include a lot of moves that tiger is going to practically
>play when it is out of book at blitz time control.
>
>Using the blitz book at blitz is logical because you save time.
>Using the blitz book at tournament time control is not a good idea because the
>demage from worse quality of moves is going to be bigger than the time that you
>save.
>
>You can look at a big database of games and analyze every position from these
>games that is relevant for tigger for few seconds for the blitz book(there are
>unrelevant positions for example if tiger's book does not include 1.e4 a6 the
>position after 1.e4 a6 is not relevant even if it was played in the games from
>the database.
>
>If you analyze 1000,000 positions and everyone of them is analyzed for 10
>seconds then you have 1000,000 new positions in your book.
>
>The time that is needed is 10,000,000 seconds of computrer time that is some
>month of computer time and the result is that you get an improveent in blitz
>that is counter productive at tournament time control.
>
>I can think of more ideas and I do not believe that the best evaluation and
>search rules for blitz is also the best evaluation for tournament time control.
>
>Uri
Right. You can also turn off the evaluation function as soon as the ply depth is
above a given depth.
This way the program will start playing worse when you give it more time.
That's not the kind of games I had in mind.
Christophe
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.