Author: Don Dailey
Date: 13:39:00 12/03/97
Go up one level in this thread
On December 03, 1997 at 14:00:50, Uwe Immel wrote: >Maybe you find interesting, below a short statement about the WMCC and >the dissapointing results of Fritz 5. F. Friedel is a member of the >chessbase company and also one of the editors of a german computerchess >magazin. > >Junior triumphs in Paris >A multimedia report by Frederic Friedel > >The 1997 World Microcomputer Chess Championship was held at the end of >October in a magnificent conference room of the Paris Stock Exchange, >the "Palais de la Bourse", which was built at the command of Napoleon >Bonaparte in the early 1800s. The Championship was organised by the >International Computer Chess Association (ICCA) and sponsored by the SBF >- Bourse de Paris, AMD, and Titus Software, a company which produces >Nintendo 64 software modules and a commercial chess program. > >There were 34 participants at the Championship. The rules allow for >programs running on a single commercially available microprocessor, >excluding machines like Deep Blue and giant mainframes. The hardware >used in Paris pushed the definition to the extreme. AMD was a sponsor >and had promised to provide the latest 233 MHz AMD K6 machines, but were >unable to do so at the beginning of the tournament. Some of the >participants had to make do with 200 MHz machines and were given a time >bonus of ten minutes per game to compensate for not getting the promised >machines. > >All this was peanuts compared to what DEC and a company called Kyrotech >supplied to some of the participants. They were experimental machines >with the Alpha processors cooled to minus 50 degrees Celsius with the >aid of freon, which took the machines up to 767 MHz. The machines were >maintained during the entire by a technician named Jeff, who was >supported by a cooling expert named Rob. > >The use of exotic hardware was not greeted with general enthusiasm. Many >of the participants had switched to 300 MHz Pentium IIs when they heard >of the hardware escalation. But some had not been able to make >arrangements in time. The Israelis, with their program Junior, had to >start with the slow 200 MHz AMDs, until Amir Ban called IBM in Paris and >talked them into bringing him a 300 MHz P2 into the tournament hall. > >The bad news > >The bad news (from our point of view) is that Fritz fared miserably in >this tournament. When playing in competition with other programs Fritz >has always had problems with it's openings book. In the past the books >paid little attention to computer tournaments and the ability of Fritz >to handle the positions that resulted. ChessBase always tried to provide >large, instructive and entertaining books for the users, and these did >not work against other computers. > >In Fritz5 our programmers retooled the entire openings book conception, >allowing very large books to be tuned by playing programs against each >other and evaluating the results. In preparation for Paris Fritz5 played >something like 1500 rapid chess games against other top programs. The >results were automatically incorporated in the openings book. The >success rate of the program in these internal tournaments provided ample >ground for optimism. However, in Paris Fritz hardly encountered any top >programs, and perhaps it was wrong to tune the program to rapid chess >games. In addition some modifications were made to the engine. After the >tournament it was discovered that in certain critical positions the >regular sales version makes clearly better decisions. > >One of the conclusions that has to be drawn is that in a tournament >against middle-class opponents an openings book consisting of 7.5 >million positions can prove to be a liability. It is probably better to >leave the book earlier with normal positions and allow the engine to >unfold its power. Naturally the final result was a big disappointment, >especially when you see the many simple draws against weaker opposition. >If you want to be a top program you have to win these games. Hi, I wouldn't attach too much significance to a single bad result. It will happen to everyone. Can you say with any certainty that the book really got you into trouble? Fritz is strong and this was certainly just a result of random chance. I doubt there were enough games to draw a really strong conclusion about this. Also, even if the games were definitely book problems, the same reasoning will apply, your sample of games is incredibly low. It could very well be your book is excellent and you were unlucky to get into one of the few bad book lines in your book. I have noticed, in my own case, that we take 1 tournament game 100 times more seriously than any of the thousands we play at home. A lost game or draw gets blown completely out of proportion. Whatever caused us to lose will become our single greatest enemy and everyone will rush to point out the problem (but rarely will anyone suggest a solution.) We try to identify our biggest problems, fix or minimize them one at a time and move on to the next one. We do not try too hard to fix a problem unless we've identified it as a big one (or it's pretty easy to fix.) Our list is very long! I am interested in your book tuning methodology. Can you elaborate? Does this one tournament convince you it is no good or do you have other evidence? Don
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.