Author: Tord Romstad
Date: 06:14:49 08/25/05
Go up one level in this thread
Hi Rolf, I'm going on holiday in a couple of days and probably won't have time to continue this debate, but at least I'll give you this reply: On August 24, 2005 at 18:23:01, Rolf Tueschen wrote: >My English is so limited that the intention is always misunderstood. Here you >thought I could really have meant to nag you with your lack of big results but I >on the other hand wanted to know the real reasons for the apparent >contradiction. If you've implemented all the known features and tricks, why >should that be still insufficient to give you a strong program? Is it really the >way you are writing, the little bugs or such things? This is hard to write to because the situation you describe is the diametrically opposite of my own perspective. I think my program is surprisingly strong, and I wonder why it performs as well as it does. The CEGT list linked to elsewhere in this thread is an old version. A newer version can be found here: http://www.husvankempen.de/nunn/rangliste.html At the time of writing, my program is number 17 with a rating 2610. If we don't count multiple versions of the same programs, it's number 14. Among the free programs, it is number 5, and among the open source programs, it's number 2. You could argue that 213 games is still not enough, but the older version 0.2.4 is not that far behind on the list (with 833 games). Considering that my program has little knowledge and is very slow, I think it is really strange that it performs so well. The only explanation I can find is that it has much fewer bugs than most other programs. It is also not quite correct that I have been working on my program since more than ten years, as you write elsewhere in the thread. It is true that I made some clumsy and buggy experiments in computer chess as early as the end of the 90s, but I didn't really know how to program at the time and never wrote a complete, working program. I only started working seriously on chess programming in 2003, when I returned to the university after a few years in the industry. My current program is almost exactly one year old. Finally, you are of course correct that many other engines improve faster than mine. There are also many that improve more slowly. I don't care. I'm not a competitive person, and I'm not in the habit of comparing myself or my achievements to those of other people. My motivation for doing chess programming is to try to make some tiny contributions to the continued improvement in the state of the art in computer chess. I regard computer chess as a community effort rather than a competition, and I am happy to see that the level of strength (especially among the amateurs) has increased enormously over the last few years. >But let me also give you a real science question and I am thankful that you are >talking with me. I understood your arguments from a non-technical side. Now the >important question: are you seriously implying that in computerchess there is no >room for always NEW inventions say like the famous Nullmove implementation? I >ask that because I read you as if all the stuff implementated you had no big >chance to invent a wheel-like new feature. Why is it so? Why should it be this >way? Are you sure it is as you assumed? I haven't claimed such a thing at all. It is perfectly possible that there are some new, revolutionary improvements waiting to be found. Finding them isn't easy, however. The beginning chess programmer has two options. He can write a conventional chess program using the published techniques, which (with enough patience) is guaranteed to result in a strong program. The alternative is to try something completely different from what everybody else is doing, which is *almost* guaranteed to result in a complete fiasco, but with a tiny chance of finding some revolutionary breakthrough. The majority of programmers will choose the easy way. For most people, working moderately hard on something that will almost certainly become a strong program is much more attractive than working very hard on something which is not likely to ever play chess well. Fortunately, there are still a few brave programmers who choose the hard way, most notably Steven Edwards. I think he will fail, but I hope he'll succeed. Nothing would make me happier than seeing Symbolic crush everybody else. >I for one come really from the outside. I wonder why you and all your collegues >are so boring and fixed on the hardware stuff while the inventions of new >features seem to be a waste of time. Is it so? I can't speak for other programmers, but personally I have very little interest in hardware stuff. >Yes, I am very naive on this topic of cloning because I don't even know how to >program. But could you perhaps become familiar with my idea that programming >creations should contain more original stuff than only copies and a little fine >tuning in style? If you can't create new features which then constitute a new >period or paradigma why do you participate at all in this race? Always copying >all the older stuff? In that meaning Peter is correct with his guess that in my >eyes certainly 80% of the ranking he gave in the other posting are just boring >clones. How boring! Chess programming using conventional methods is definitely closer to engineering than to science or art. Perhaps you don't like this fact, but it remains a fact. Tord
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.