Author: Alessio Iacovoni
Date: 23:03:54 10/22/98
Go up one level in this thread
On October 22, 1998 at 23:01:37, Komputer Korner wrote: >On October 22, 1998 at 04:47:41, Alessio Iacovoni wrote: > >>On October 22, 1998 at 04:19:40, blass uri wrote: >> >>> >>>On October 22, 1998 at 04:04:33, Alessio Iacovoni wrote: >>> >>>>On October 22, 1998 at 03:31:10, blass uri wrote: >>>> >>>>> >>>>>On October 22, 1998 at 03:22:57, Alessio Iacovoni wrote: >>>>> >>>>>>I just finished my tests with the engine-engine function of Junior. This time I >>>>>>was careful to reduce hash sizes in order to avoid any mistakes from possible >>>>>>delay due to the hard disk grinding (I believe this was the reason for those >>>>>>problems in the previous tests). >>>>>> > >2 points. > >A LOT OF PLAYERS use the analysis features for overnight analysis or for 10 >minutes or more on one position. For that; games at 40/2 are a much better >indicator of strength. 2nd point is that the user can never be sure that the >program is or isn't optimized for the tournament book that comes with the >program. Indeed in most cases it IS optimized by the programmer to fit the >engine's style. Opening repertoires are part of the equation of a chess player. >Computer programs are no different. >-- Point 1 is ok. But I don't agree with point 2. I don't know if I remember wrong but didn't chessbase for example claim that it's tournament books were just made from a wide collection of grandmasterlevel games? Apart from choosing open lines instead of closed ones I cannot frankly think of any specific optimization applied to engine X instead of to engine y or z. As a matter of fact, from what I've read so far, major concern is in having a new opening book during tournament that was not "autoplayed" previously and tuned to other engines... (see recent posts). Basically the work done is in finding lines that lead to the greater amount of wins, with greater elo, in open positions, etc etc.. but that has nothing to do with "tailoring" one's opening book to his specific engine. >Komputer Korner >>>>>>Processor: AMD 200 MMX >>>>>>Cache: 8 M >>>>>>Total games:32 >>>>>>Time control: 10 minutes >>>>>>Book settings: Same openings played by both sides >>>>>>Book: Fritz5 >>>>>>Result: >>>>>> >>>>>>Hiarcs 6: 10 >>>>>>Junior 5: 10 >>>>>>Draws: 12 >>>>>> >>>>>>Just from these first results it is interesting to notice that when both sides >>>>>>play the same opening (eliminating any randomness), playing strength seems to >>>>>>even out incredibly. I wonder if test are carried out in the same way by SSDF. >>>>>>If not, why? >>>>> >>>>>not by the same way. >>>>> >>>>>1)They give every program its opening book because they want to test the level >>>>>of the program(good opening book is a part of it) >>>> >>>>As I have already stated in the past I don't agree with this... When talking >>>>about the strength of a program people generally (aside for experts) do not >>>>refer to the book but to the engine...Does SSDF explicitly state that program A >>>>+ book A ranks higher than program B + book B? If not it leads people to think >>>>that it it the engine alone that is stronger, which is not the case (see any >>>>thread on computer.chess). That I know of the SSDF lists do not mention at all >>>>the book used in listing the strength of the programs. >>> >>>They assume that it is obvious that they use the books that are sold with the >>>programs. >>> >> >>>I saw games and I know that they use the books of the programs. >>> >>>Sometimes the books are counter productive. >>>I saw a game between fritz5(Pentium200) and Mchess6(P90) that was a draw by >>>book(firtz5 was white). >>> >>>I think that fritz5 had good chances to win without book. >>>, >>>>> >>>>>2) they do not do engine v engine but play with 2 computers so every program can >>>>>"think" in the opponent's time. >>>>> >>>> >>>>Ok.. I know that. Separate computers is better. >>>> >>>>>3)They play 2 hours per 40 moves and not 10 minutes for all the game. >>>>> >>>> >>>>Why? Is that the time control usually adopted by the people that buy computer >>>>programs (WOW 2 hours/40 moves seems quite boring to play against a computer). >>>>Also.. can the ranking of a computer at 2h/40 moves say anything about it's >>>>strength at 30 minutes per game? or 10? >>> >>>I am interested in longer time control because I use the computer to help me in >>>my correspondence games. >>> >> >>Come on Uri.. you're not the "average" 90% i was referring to. >> >>>Uri >>>> >>>> >>>>>Uri
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.