Author: Bob Durrett
Date: 17:00:11 12/27/03
Go up one level in this thread
On December 26, 2003 at 21:24:06, Uri Blass wrote: >On December 26, 2003 at 21:08:08, Mike Byrne wrote: > >>On December 26, 2003 at 19:22:45, Bob Durrett wrote: >> >>>On December 26, 2003 at 16:17:38, Mike Byrne wrote: >>> >>>>On December 26, 2003 at 15:34:43, Darren Rushton wrote: >>>> >>>>>>Actually what happens, is the 366 is SLOW. And I mean SLOW. >>>>> >>>>>I don't intend to be controversial here, but the conclusion I draw from your >>>>>results is that Shredder 7 is such a brilliant program it is almost a match for >>>>>the one of the better amateur programs on hardware that's almost 10 times >>>>>slower. >>>>> >>>>>Regards, >>>>> >>>>>Darren >>>> >>>>It's not controversial and it's not new. In fact, these results could have been >>>>predicted. Anytime you have a rating difference of around 200 points (SSDF >>>>Shredder 2812, Crafty 2615) running on the same hardware, you would expect fair >>>>match would be on a somewhere where the weaker engine need a 8x to 10x speedup. >>>> It is often believed that a doubling of speed increase strength of around 50 >>>>elo points (some feel more, others less) especially in computer vs computer >>>>play. 200 point rating differential would indicate: >>>> >>>>2^3=8 >>>>50x2x2=200 points >>>> >>>>therefore one would assume that it would require ~approx 8x machine would for >>>>the weaker engine to catch up... >>>> >>>>SSDF >>>> Rating + - Games Won Av.opp >>>>1 Shredder 7.04 UCI 256MB Athlon 1200 MHz 2812 28 -26 781 75% 2623 >>>>30 Crafty 18.12/CB 256MB Athlon 1200 MHz 2615 27 -27 647 52% 2604 >>>> >>>>In my view, Shredder 7.0.4 is the stongest engine available and I have seen >>>>nothing to indicate that it should change. >>> >>>As far as I can tell so far, Shredder seems to give better results. Now if the >>>author of shredder will just fix the bugs and fix it so that the PV is "high >>>quality," Shredder will be the "best" chess-playing program. >>> >>>Bob D. >>> >> >> >>Shredder is the best "playing" program (imo), the pv feature does not >> ------- >>impact its play -- I suspect the PV in any computer program is always suspect, >>for serious analaysis - move it one move at a time - then analyze for 3 minutes >>(or whatever) andthen move on ...just to look at any pv from most programs will >>get you in trouble... >> >>I will say that the Shredder PV is more suspect than most, and perhaps for >>someone like yourself, it is not the best overall program. >> ---- > >I can explain the problem that I see with wrong pv. >The pv should be the line that the program starts to search when it does a new >iteration. > >If shredder starts to search from it's pv then it means that it has often poor >order of moves in the start of the iteration. > >It is known that programs can play better if they improve the order of moves so >it means that shredder can play better or does not give information in the pv >about the line that it starts to search. > >I do not expect pv to give the best moves but I expect it not to give stupid >blunders when even searching to depth 1 can avoid them(except maybe the last >plies of the pv because the pv should be based on exact results of search to >reduced depth). > >Uri That is most interesting. Perhaps "PV" has different meanings and uses depending on the programmer. Each programmer may have a unique way to obtain the "PV" too. There are two types of people who care about PV, IMHO. They are those who are trying to evaluate the performance of the software, and those who use these chess-playing programs for other purposes. Of those trying to evaluate the performance of the software, the programmer will perhaps care the most. If I were that programmer I would make the program produce outputs which would help me evaluate my software, at least during the development phase. The "PV" may be one of these outputs. Perhaps it is the usual practice for programmers to cause the software to produce many special outputs [and maybe putting data into memory for later evaluation] during the development and testing of the programs. Then, after the program is published, or sold for use, then maybe most or all of these special outputs might be disabled. In any case, if the GUI cannot process the data, then it really doesn't matter what the engine outputs to a GUI. Maybe the PV produced by Shredder is still useful to it's programmer even if other users cannot use the PV data? Incidentally, will the nature of the PV depend on which GUI [or UI] is used? In any case, the programmer of a commercial program should make extra effort to assure that the PV data is useful to the customers. Otherwise, it should not be provided to the user. Bob D.
This page took 0.01 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.