Author: blass uri
Date: 00:42:02 06/14/00
Go up one level in this thread
On June 13, 2000 at 20:21:25, Dann Corbit wrote: >On June 13, 2000 at 19:57:39, Mogens Larsen wrote: >>On June 13, 2000 at 19:18:45, Dann Corbit wrote: >>> And even the opening database data (while - admittedly - more important than >>>endgame tablebase data) is not the achilles' heel. If you want to strike at the >>>heart of a chess program, simply remove the data from the eval function. Now >>>we'll see who plays crappy chess. Essentially, what you will have is my >>>retarded move generator chess program. The GM's won't have much problem with >>>that, but neither will anyone else for that matter. >> >>You still don't get it. It's not a question of crippling a chess program, not as >>far as I'm concerned, and it involves other considerations than the simple ones >>you present. Of course a human brain is responsible for the code involved in a >>chess program and the same thing applies for opening books and endgames tables, >>so it has no bearing on the discussion at hand whatsoever. > >The evaluation terms -- where did they come from? Ask Dr. Hyatt. I have seen >many comments like "Roman said to do this" or "GM so-and-so suggested that..." > >Every facet of computer chess programs is a merger of data gleaned from chess >experts with algorithms written by computer experts. > >>Opening books and >>endgame tables are simply not a product of the programmer and rarely something >>produced by the program itself. > >This is completely and in all ways irrelevant. The *EXACT SAME THING* is true >with the data used by the evaluation function. And the endgame database files >are written (generally speaking) by pure programmer types and so are the least >"chess genius" part of the system and (ironically) the only part that is >perfect! > >>A chess program should play chess on its own >>terms, not through more or less random exterior additions and attachments. > >You don't understand how programs work very well. A program operates on data. >Without those "additons and attachments" there is no program. Period. The >opening book data is *not different* than the eval data or the endgame data or >any other data. It was produced by someone who knew what they were doing. > >>You >>might say it's a question of existentialism. What constitutes a genuine computer >>chess program? > >Data Structures + Algorithms = Programs. >Wirth knew what he was talking about with that one. > >I write business systems. In fact, if you are using a Windows front end to type >in messages, I probably wrote one of the software tools used to process the >purchase and manufacturing transactions [that's right, I have been sleeping with >the enemy]. >;-) > >Anyway, I don't understand the manufacturing data as well as the manufacturers. >I don't understand the purchase transaction as well as the POS folks. But with >their expert data and giving me the instructions I needed to process the >transaction, voilla -- 40 million units get processed in a couple months. > >There is a synergy between the data that gets processed and the algorithms that >process them. The programmer needs to understand the data. But he will never >understand it as well as the people who generate the data and operate on it >every day. However, as long as sufficient system information is gathered to >produce results that meet requirements [cough -- assuming the requirements >reflect the actual need of the end-users] then the programs will produce the >required results. > >A chess program is just another instance of this. You go and gather data from >*true* experts in the field (typically chess players, chess books, and so on). >This data is coupled with state of the art algorithms to produce chess programs. > The chess programmer is rarely a GM (though Vincent and Gullydeckel's author >Martin Borriss are clearly expert players). Even so, in that case, the author >of the program is the source of the data (just another source, like any other). > >The programmer almost never invents the algorithms he/she uses either. Will the >move ordering sort be invented by the programmer? There is only one inventor of >NULL move and only one inventor of Alpha-Beta search, yet almost all chess >programs use these features. I think that it is a mistake not to try to invent algorithms because I do not believe that the algorithms are the best. For example I believe that programmers should try to invent selective search ideas and try to define illogical moves but it seems that nobody tries to do it and they continue to use the same brute force. Good humans sometimes can see more than computers in tactics because they use selective search and they know that moves are illogical without analyzing. It is not easy to define selective search rules without pruning logical moves but the problem is that I believe that most programmers do not even try. The way to go should be to define some rules and check for the reason that they fail and change the rules again and again until they do not fail. Uri
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.