Author: blass uri
Date: 03:34:26 04/21/98
Go up one level in this thread
On April 20, 1998 at 19:47:27, Thorsten Czub wrote: >On April 20, 1998 at 17:39:39, Tord Romstad wrote: >>Try this position (white to move): >> >> +---+---+---+---+---+---+---+---+ >> 8 | | *K| *B| *R| | | | *R| >> +---+---+---+---+---+---+---+---+ >> 7 | *P| *P| | | | | | | >> +---+---+---+---+---+---+---+---+ >> 6 | | | | | | | | | >> +---+---+---+---+---+---+---+---+ >> 5 | P | | *N| | | *P| *P| *Q| >> +---+---+---+---+---+---+---+---+ >> 4 | | | N | | | | *N| | >> +---+---+---+---+---+---+---+---+ >> 3 | R | | | | Q | | P | | >> +---+---+---+---+---+---+---+---+ >> 2 | | | | B | | P | | | >> +---+---+---+---+---+---+---+---+ >> 1 | | | R | | | B | K | | >> +---+---+---+---+---+---+---+---+ >> a b c d e f g h >> >>Genius 3.0 and older does not find the mate (1.Qf4+ is mate in 21 plies) >>before search depth 9/21. However, Genius 4 and 5 finds the mate at much >>lower search depths. Because of this (and other positions) I believe >>that >>your conclusion is correct for old versions of Genius, but that the most >>recent versions extend some lines beyond the 12 plies of selective >>search. >> >>Tord > >Explanation in my opinion: >historical business reasons. >In germany there was a heavy fight in seeling more units between Genius >and Fritz. >Fritz-company always claimed that fritz is a good blitz-program due to >advertising campaigns where Fritz had to play against Kasparov and other >players in blitz or active chess. Therefore the traditional good blitz >program Genius had to be TUNED to stand Fritz or get as good results in >tactical test-suites like Fritz. >In the end all this TUNING of Genius made Genius weaker imo. >Genius lost the race with Fritz. > >I guess Richard is a very nice person. He cannot write a >massive-attacking-monster killer program like others can create. >He is more a nice guy. I do not think a nice guy cannot write what you call "a massive-attacking-monster killer program" I don't say about any program it is a killer program >Genius always played like Richard behaved, in the end, it looks Richard >lost the race because he was unable to defend against people attacking >genius. I think there are other reasons that Genius does not play well. it's evaluation function is sensitive to the root position(at least for Genius3) It was a good idea for slow machines because in this way it can analyze more positions but is not a good idea for fast machines. I know fritz5 also suffers from this problem. from my experience Junior4.6 has not this problem. I want to know what are the programs that has an evaluating function that is not sensitive to the root position. >He was not able to DEFEND or stand the attacks. >Ossi always tried to force him to stand them, and helped, but you cannot >make a samurei out of a pacifist. >Fritz on the other hand, armed with big hash, power-books and a very >tough team (the genius team was IMO too much directed by Ossi. Team work >is not ONE master and a few workers, team work is MORE. In the end this >team was not good enough) I am not sure if power book realy help fritz It lost against junior4.6 in the last game in paris mainly because of some power book. I am not sure if big hash tables realy help fritz5(20Mbytes are much better than 2Mbytes but I do not think there is a big differnce between 20Mbytes and 100Mbytes I saw from my experience that fritz5 increase the selective search depth with big hash tables but I think it is more important to increase the brute force depth and sometimes fritz5 is even slower with big hash tables in reaching the same brute force depth. maybe the optimal hash table for fritz5 is something like 60Mbytes and one of the reasons fritz5 failed in paris was that it used too much hash table. I don't think so but it is a conjecture that should be checked. > >The race is over. There is no more genius. >They said they would tranfer it from 16 bit assembler into 32 bit >assembler. >But I don't believe it takes so much time. I have the feeling the >relationship between ossi and richard is broken and for this reason >there is no more Genius follower. But this is only my explanation. > >I still think tuning genius into a wild animal/beast was wrong. >They would have done better to increase genius strengths instead of >trying to exchange genius-weaknesses by implementing extensions for >tactict and changing its style. I do not agree. tactic is very important in chess and I do not see contradiction between increase genius strengths and exchange genius weaknesses. the most important thing in order to do a better program is to exchange weaknesses of a program. >It played boring. This was >system-imanent. YOu cannot change this the way they tried to open the >playing style in Genius4. Genius5 took back those efforts. > >Will there ever be another genius. >I hope so. But the chances are low, or ? >Lets keep the fingers crossed.
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.