Author: Dann Corbit
Date: 21:12:12 01/13/06
Go up one level in this thread
On January 13, 2006 at 23:32:04, Dann Corbit wrote: >On January 13, 2006 at 23:08:25, Uri Blass wrote: >>On January 13, 2006 at 19:20:04, Dann Corbit wrote: >>>On January 13, 2006 at 18:28:52, Uri Blass wrote: >>>>On January 13, 2006 at 18:12:17, Dann Corbit wrote: >>>>>On January 13, 2006 at 17:32:10, Uri Blass wrote: >>>>>>On January 13, 2006 at 16:54:55, Reinhard Scharnagl wrote: >>>>>>>On January 13, 2006 at 15:58:57, Robert Hollay wrote: >>>>>>>Hi Robert, >>>>>>> >>>>>>>>Shareware. Test SMIRF free until 2006-01-31. >>>>>>> >>>>>>>the beta still shows to be intended to become shareware. But I am not sure >>>>>>>about that. There is no big interest facing those Rybka or top-engine hype. >>>>>>>About since a year there repeatedly have been unrestricted beta versions, >>>>>>>limited in lifetime to secure vanishing of old beta prereleases. But the echo >>>>>>>has been very small and only one person asked, how to purchase a full version. >>>>>>> >>>>>>>>What will happen after that date? >>>>>>> >>>>>>>The beta version still could be tested, but the program would answer nearly >>>>>>>immediately then, not reaching higher thinking levels, nag screens included. >>>>>>> >>>>>>>>I looked around the SMIRF site, but didn't quite understand it's licencing >>>>>>>>model. Part of the text is written in German, especially the shareware licence. >>>>>>>>What should we do to get it permanently? >>>>>>> >>>>>>>A sold / donated version would have texts also translated into English. >>>>>>> >>>>>>>SMIRF is now in a stage, where it should be rewritten completely to dense its >>>>>>>data structure and to incoorporate made experiences. I am against open source >>>>>>>programs for end-user applications like a chess engine. Arguing for that had >>>>>>>raised a banning from the most visited german chess forum. I think, that >>>>>>>freeware chess programs are dumping the whole scene, killing any respect and >>>>>>>refund of creative programmers. >>>>>> >>>>>>It is correct that free chess programs clearly make it harder for programmers to >>>>>>make money from chess engines but it seems to be also correct that free chess >>>>>>programs help to improve the level of the best programs. >>>>> >>>>>I doubt this very much. >>>>> >>>>>There would never have been any money in mediocre chess engines with or without >>>>>free ones. I believe that free engines stimulate the market. In fact, I am >>>>>pretty sure I would never have bought any professional engines if it were not >>>>>for the free ones to get me interested in them. >>>>> >>>>>Suppose that there are no free engines. Your program will still have to beat >>>>>Junior, Shredder, Fritz, CM, etc. to be interesting as far as strength goes. >>>>> >>>>>And if you try to sell on features you will need a snazzy interface and a good >>>>>database and hundreds of thousands of dollars for marketing. >>>>> >>>>>I think that the engines like Deep Sjeng, Ruffian, Rybka, Ktulu, etc. would not >>>>>sell at all if it were not for the hobbyist market created by people who cruise >>>>>forums like this one. >>>> >>>>I disagree here. >>>> >>>>Deep Sjeng and ruffian and Ktulu probably do not have many buyers because >>>>correspondence players who want to buy only the best 2 engines do not buy it >>>>when rybka may have more buyers than all of Deep Sjeng,Ruffian and Ktulu. >>> >>>So how will a weaker amateur engine figure into this mix? How will the >>>existance of these engines cause problems? >>> >>>I doubt if any of the non-Chessbase non-ChessAssistant non-Chessmaster engines >>>have sold more than 3000 or will sell more than 10,000 in their lifetimes. In >>>any case, the lack of sales has nothing to do with amateur engines and >>>everything to do with advertizeing, shelf space, etc. >>> >>>> I get if you ask Lokasoft, almost all of his sales will >>>>>go to people who he recognized the names of (just a guess). >>>>> >>>>>>There was a long time with little improvement when shredder dominated and top >>>>>>programmers could not improve their program even by 50 elo per year and suddenly >>>>>>we find some programs that are better than shredder(rybka,commercial fruit and >>>>>>fritz and probably hiarcs and the toga1.1 that is based on fruit). >>>>> >>>>>The improvement always goes in jumps and starts. The introduction of Junior was >>>>>like that. Junior came out of nowhere as an amateur engine to be the best in >>>>>the world. >>>> >>>>Junior never leaded the ssdf and I know no rating list with many games when it >>>>leaded. >>> >>>Incorrect. I think it may have won more than once, but here is one of its >>>winning lists: >>>http://www.geocities.com/CapeCanaveral/Launchpad/2640/ssdf/2000/ssdf0001.htm >> >>Ok I apologize and I was wrong here but it leaded it only for a short time and >>by a small difference. >> >>Rybka is leading rating lists by bigger difference. If we compare the 32 bit Rybka with the 32 bit Fruit from here: http://www.husvankempen.de/nunn/cegtrating4040best.html We see a difference of 38 Elo, about the same as the Fruit lead over Fritz 9 of 33 Elo on the current SSDF list. Historically, many of the Elo leads have been small in the list and I did see one tie. But I also saw the following differences in the lists between 1996 and 2000: 23 Oct 1996 +41 Elo lead for Rebel 8 24 May 1997 +71 Elo lead for Hiarcs 6 22 Feb 1998 +55 Elo lead for Fritz 5 27 Oct 2000 +38 Elo lead for Fritz 6 The +24 Elo lead for Junior over Chess Tiger 12 in 30 Jan 2000 was well above average, and Junior was also second in one of the lists that I saw. I am still looking for a site with more historical data on the SSDF lists. I think it will be an interesting area to do some statistics. I am especially interested in the intervening lists, because I can't find data for years between 2000 and now. Probably, I could find most of them by searching the CCC archives (but the search stuff is mostly broken). >Rybka is in second place: >http://www.husvankempen.de/nunn/cegtrating4040all.html > >In case you want to object because of SMP, I would argue that SMP is an advance >just like better search or better eval. It is not the fault of SMK that Rybka >and Fruit can't do it yet. > >>>It also won the world championship 3 times: >>>http://www.mark-weeks.com/chess/wcc-comp.htm >> >>I know it but the point is that the improvement now is bigger than the >>improvement in previous years. > >I have not done any graphs, but I do agree that there may have been a bigger >spike than usual recently. I interpret this as really smart people getting >interested in computer chess. The same sort of thing has happened in the past. > >I think that the startling improvements are due to smart people with good ideas >and enough drive to implement them. > >>>> It was fairly similar for Chess Tiger (which occupied the SSDF top >>>>>spot for a short while). So Rybka is not an exception. I guess that a careful >>>>>examination of all the big movers and shakers will show that Rybka is the rule. >>>> >>>>I think that rybka is clearly an exception if you compare it to programs of >>>>2003-2005 >>> >>>Fruit. >> >>No >> >>Fruit commercial version was less than 50 elo better than shredder >>When Rybka64 bit has bigger difference from the second place based on rating >>lists. > >It is still relative. Fruit was a blast from a cannot. Rybka was too, with a >little bigger warhead. But Fabien may have a bigger warhead on the shelf. > >Some of the others may also innovate. The Ruffian author was very quiet for a >while and suddenly unleashed his creation. There may be another phase of that. > >I think that some of the most capable authors just don't put enough effort into >it to become overpowering. For instance, the author of Gullydeckel is a very >good programmer and a very good chess player. I think he could add 1000 Elo to >his program (literally) but he is too busy doing real work. Simlarly, the >author of Amy is quite a brilliant person and could advance with big effort. >But again, I think that hobby time is lacking. > >To write a great chess engine requires two things: >Great ability >Great effort > >What we are seeing is those things in combination. > >You also need to be a little crazy (since the time spent writing the chess >engine could be spent -- for instance -- making money or having some recreation >or whatever). > >Wait, that's three things. > >And did I mention an almost fanatical devotion to the pope? > >>>>There was only little improvement of less than 50 elo from shredder7.04 to >>>>shredder9 when the improvement from shredder9 to rybka1.2 is probably going to >>>>be more than 100 elo. >>> >>>Fruit from 0.98 to 2.2.1 is about 300 Elo in less than 2 years. >> >>This is irrelevant because I compare with the top programs and not with an old >>version. > >It might be irrelevant, but it is also pretty incredible. It was a hot >steam-roller right out of the box. And in the short time span, he turned it >into a towering juggernaut. > >>>I guess that in 2 years more Fabien can add 150 more if he wants to put in the >>>effort. >> >>I agree but my point is not about future fruit but about comparing the progress >>today and in the past. >> >>Today it is clear that there is a bigger progress. > >It's not clear to me. I think that the biggest jump ever was pre-deep blue to >deep blue. Everything else has been a tiny blip by comparison. But that's not >a fair comparison because they had a huge and talented team with a huge budget >and spent huge time and resources. Nobody can compete with that. Except maybe >Hydra. Not sure about Hydra, though. They have already spent a big pile, >that's clear.
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.