Author: Serge Desmarais
Date: 20:45:44 09/08/98
Go up one level in this thread
On September 08, 1998 at 17:34:56, Don Dailey wrote: >On September 08, 1998 at 13:19:24, Moritz Berger wrote: > >>On September 08, 1998 at 12:24:07, Don Dailey wrote: >> >>>I think it is still fundamentally superior to the other programs. It >>>may not actually be the very strongest currently, but this may be because >>>Richard has not made any substantial effort to stay ahead and I also >>>don't think the book is engineered as well as the top contenders, which >>>could be a big part of the reason his program is not dominant right >>>at the moment. >>> >>>- Don >> >>I challenge your reasoning above because several facts seem to contradict your >>statement about the book being to blame at large for Genius not being >>"dominant": >> >>(1) #1 at the SSDF is still Fritz 5, which uses a GM database without *any* >>modifications or tuning (PowerBook). So your statement about "book is [not] >>engineered as well as the top contenders", given the lack of engineering in >>Fritz' case, seems funny to me. >> >>(2) Genius 3 had one of the best opening books of its time. Do you think that >>Richard Lang (or Ossi Weiner or whoever) replaced it with something inferior? >>Why should they do this? >> >>(3) Try it out yourself: Using the G3 or G5 or Genius tournament book (the >>infamous 1.c4-only book) doesn't have any measurable effect on results vs. other >>programs or humans (i.e. all books I mentioned fit Genius' style equally well). >> >>(4) Test suites don't have Genius at the top, in fact it fails in some positions >>that all other contemporary 2500+ SSDF programs solve in a couple of minutes. >> >>(5) "Dominant" means at the very least 50 ELO, more likely 100 ELO points >>stronger. Hiarcs 7, Rebel 10, Junior 5, Fritz 5 score (or will score with high >>probability) about 50-100 ELO better than Genius 5 at the SSDF (and probably >>anywhere else, for that matter). >> >>For Genius to become "dominant" at some other "moment" in time, it would need to >>gain a minimum of 150+ rating points with the next release, something which >>seems slightly unlikely given the historically small improvements e.g. from >>Genius 2 to Genius 5. >> >>Also, the reign of Genius already ended 5 years ago with the advent of Hiarcs 3 >>and Rebel 6. The era of dominance was most significant with dedicated Mephisto >>machines and maybe the very first Genius releases for the PC (1 and 2), but >>after that true improvements happened elsewhere, IMHO. >> >> >>Moritz > >Hi Moritz, > >Thank you for challenging me on this one. I admit a lot of this was >simply my own subjective impression and also a lot of speculation. > >I agree that at one time Genius opening book was quite good. I >currently believe it to be less so, not because they degraded it but >because others have caught up and surpassed. I think this explains >some of the strength difference but I admit I could be wrong, what >I said was speculation on my part. > >Call me superstitious, but I don't trust ratings at the very top >end and at the very low end of any given rating pool. Based on >many studies I have personally done with various versions of my >own program sometimes combined with other programs, the phenomenon >I always see is the the top rated guy is significantly overrated. >The rating of this guy will always drop (when using some intermediate >level to standardize against) when stronger individuals are added >to the top and they in turn will become overrated. I believe this >phenomenon exists in the SSDF and any other rating list you present >to me. Fritz may be strongest but I'll bet it's not by the margin >it appears to be by the numbers. Therefore, I strongly suspect that >not so much difference exists between Genius and the top guy. But >I did admit Genius is currently not the very best which is in harmony >with your assertion. > >I still stand by my statement that Richard has made no effort to >stay ahead of the rest. His program at one time was so far ahead >of the pack that he never needed to worry about this, but he does >now, only I don't know if he cares much any more. > >- Don Maybe he didn't find a way, YET, to significantly improve his program? It is tough to improve something that is very good? On the Genius home page at http://www.computerchess.com/chessgenius/home_e.html , it says/said that Richard Land has or is rewritting all his code in 32 bit optimized code but that the improvement over Genius 5 is/was still quite small and so he would have to try new ideas for a sensible improvement. Serge Desmarais
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.