Author: Robert Hyatt
Date: 18:39:17 10/28/99
Go up one level in this thread
On October 28, 1999 at 16:51:56, Alexander Kure wrote: >On October 28, 1999 at 15:43:02, Ricardo Gibert wrote: > >>On October 28, 1999 at 14:50:42, Robert Hyatt wrote: >> >>>On October 28, 1999 at 14:26:57, Ricardo Gibert wrote: >>> >>>>On October 27, 1999 at 15:22:29, Alexander Kure wrote: >>>> >>>>>On October 27, 1999 at 11:55:49, Amir Ban wrote: >>>>> >>>>>> >>>>>>You noplayed the two highest rated blitz computers around (Ban & Varguz). The >>>>>>others are either hundreds of points lower or don't come any longer. >>>>>> >>>>> >>>>>As I run Varguz on my second compi I think I can add something: >>>>> >>>>>I do not care if Crafty is noplaying Varguz. I do not have a noplaylist, cause I >>>>>simply do not care. I think Crafty is an amateur program that could learn a lot >>>>>from playing the commercials. >>>>>Let me put it this way: Compared to the strong commercial programs like Junior, >>>>>Fritz, Hiarcs and Nimzo Crafty is much weaker and needs some improvements to be >>>>>a match for the above programs on equal hardware. >>>> >>>>Be fair. You are forgetting that the commercial programs do not publish their >>>>source code. Sure they are stronger, they have the crafty source code to look >>>>at. Due to this, one might say that crafty defines the dividing line between the >>>>top programs and the also rans. If your program is not at least as good as >>>>crafty, then it is not a top program. No matter how much Hyatt improves his >>>>program, he will never be able to best the top programs (for very long anyway), >>>>since he gives away all his ideas. >>>> >>>>> >>>>>Greetings Alex >>> >>> >>>And we _know_ that nobody uses those outdated ideas in Crafty, right? :) >>> >>>IE book learning, some of the evaluation terms... etc. I have been really >>>hesitant to release the new pawn majority code because no one else is doing >>>this that I can recognize, at least not doing it effectively. The minute 17.0 >>>hits the street, others will be doing it soon. >>> >>>Good or bad? >> >>Not just good, great! Crafty has doubtless elevated the playing strength of all >>chess programs (both commercial and non-) by making the source public. Alex got >>it backwards: The important point is not so much that crafty _could_ learn a lot >>from the commercial programs, but rather the commercial programs _do_ learn a >>lot from crafty. >> >>> >>>A one-way flow of information? >> >>Yes. If they want to compete with crafty on an equal footing, they should >>publish their source. >> >>> >>>Is it really worth it? > >I cannot confirm that commercial programs _do_ learn a lot from crafty's source >code. What can be learnt from it? If it weren't for the privacy of email, I could give you a quick example of how much of crafty's book learning code is in Fritz. As one example. And I don't know about the programs that didn't do it 3-4 years ago, but do do it now. > How Passed Pawns are being handled. Not as >effective as Nimzo does. I disagree there. There are some things I have been doing, particularly in v17, that I see _no_ evidence of anyone else doing. I watched Nimzo trade into a dead lost ending today on ICC. My majority/candidate code instantly said "Ok, he wants to trade pieces, lets trade _all_ pieces" because my candidates on tha a and h file are much more unpleasant for him than his two connected passers in the middle. It was right. I'm sure there are still bugs. But don't just assume that the commercial programmers have _all_ of the right answers. You would be wrong more often than right. We have a pretty good group of 'amateurs' that share lots of ideas, often in private to keep them out of 'public' eyes. Well what about Crafty's King safty - Nimzo is very bad >on this - surely I can copy from Crafty's - sounds ridiculous to me. >Crafty's opening book? Nice compilation of standard games but what about anlysis >that has never been played? Think my book does the better job. >Just to think there is a source code available for free and all people - >including the proffesional chess programmers - jump on it like thirsty people in >the desert - hm. >I also cannot confirm that there is what you call a one-way flow of information. >Ever been to a computer chess event? Both professional and amateur programmers >exchange a lot of ideas there. I have been to them since they first started holding them. In the early 70's, we all talked. In the 80's, the 'research groups' talked, the commercial programmers 'lurked'. To see this, just take a gander through the JICCA, and see how many published articles you find by commercial programmers. Prepare to look long and hard. And prepare to find only a tidbit here and there that is 5-6 years out of date. Compared to those of us doing this for fun... The gap is _huge_. > Can you image all the e-mail contact between >them? Maybe they are talking a little bit more than about the wheather? >There are many different channels of information exchange which you may not be >aware of. So asuming you never heard of anything does not imply that is does not >exist. Ever asked Chrilly Donninger about how he has implemented Null Move? Ask >him and he will tell you! maybe he will, and maybe he won't. But the usual way of dissiminating scientific information is thru public discussions (such as here) and publicly available Journals (such as the JICCA and IEEE among others). If you look thru todays posts, Christophe mentioned that he used both pvs and alpha/beta. Someone asked him "How do you choose when to use each". The answer was the very common commercial response "I don't want to reveal too much." Did Lang ever reveal the details of the search he used in Genius? etc... > >Just to put things right. I think what Bob does is wonderful for the computer >chess community. Many amateurs profit from what he has done and will profit from >what he will be doing in the future. I personally appreciate this! Not because I >will profit personally (I am not interested in that) but because with this he >keeps many people (amateurs) in the run for learning and improving their >program. > >Thx >Alex that's the only reason I do it, too. I grew up in the world when I could call Slate or Thompson and ask technical questions and get real answers. Not "I don't want to reveal that". I believe that newcomers deserve the same kind of environment that I received. I'm going to do my part to help. If the commercial guys feel ok with participating here without really participating, that's ok. It would bother me to read and use good ideas, without ever supplying anything in return. That just feels wrong. IMHO.
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.