Author: Ren Wu
Date: 18:16:26 07/30/02
Go up one level in this thread
On July 30, 2002 at 20:07:05, Vincent Diepeveen wrote: >On July 30, 2002 at 17:30:53, Ren Wu wrote: > >The strength of a game is also determined by how many professional >players a game has. We see it clearly in tennis and table tennis how >a sport from very amateuristic level has developed in a very professional >scene. This definitely is the case with chess. What is your rule to classify professional and amatuer? The number of pros are certainly help, but not a _determined_ factor. BTW, in china along, there aer easily most than 150 pros. >If i see how many nodes a second you get in chinese chess and see >the average number of possibilities (which in chess is 40 exactly >i measured with DIEP, not 35 like it is in human games), and the >huge influence of both scientists and other researchers in the >field then the facts are pretty easy. I don't know what do you mean here. >Let's be clear here. How many parallel running chinese chess programs >are there? I don't know, but certainly is more than one. But is this important? >If little, why? Especially if you consider the search depths they get, >for the top programs to play another top program, getting 2 ply more >is pretty interesting. I don't know what depth other porgram reached and their hardware. But I am not sure why that is 'pretty interesting'. I remember saw your posts long time ago, that your program was been out-searched by other programs 2-3 plies, and I did not seems surprise much. The reason is that different author have different opinion on how things should work. Some prefer more knowledge, some more prefer more speed, some prefer expensive extensions, some prefer aggrasive pruning. I respect their decisions. >Branching factor is determined by a few things > a) quality of the program > b) number of possibilities Right. >The go programmers at the mailing list always complain about branching >factor, yet my GO program has from the openings position a branching >factor of 10.0 > >Of course using nullmove. >Yet they claim way higher branching factors. I believe you are compare your effective branch factor to their branch factor. This has no use in this context. >In chess we had a few years ago the same thing. Huge branching factors. >Now we have way smaller branching factors in chess, after years of >hard work. Same to computer xiangqi. >In all ways computerchess has developed itself very professional. Claiming >there is no such a gap between chess and Xiangqi is in itself a complete >denial of the facts. computer chess is more professional developed. But that will not change the fact that best xiangqi program is at same level of best chess program. >Let me show you one reason why many game developers in the east are suffering >from problems: > - not being able to speak english Educated young generations in China is pretty good at english. They may not have enough practice on oral language, but they certainly very good at read. English is not a problem at all for them. 30 years ago, it was, it is not now. Every year there are few Chinese student got full score at TOEFL, and GRE test. >As a result of that i can't remember many serious papers regarding Xiangqi. >Please explain to me one algorithm that in generally works (both chess AND >Xiangqi are very similar, of course mating the king is completely different >in both games, so will extensions be here) for both games which has >been developed by a Xiangqi researcher? Only search-related paper is interesting enough for acadmia world. If people want to publish a paper in english, why they want choice a unformiliar game instead a known one? And do you see any professional chess programmer publish papers about their insights? Let me tell you something about xiangqi. knowledge is vitally important in xiangqi endgames, because there is no promotion. precise knowledge is needed to play every type of endgames. long-term positional sac, without it, you can't even reach the top amatuer level. There are many pruning/extending rules works for both chess and chnese chess but those problems above can not be solved at algorithm level, but rather at system level. >I'm sure if i just look to hashtable management of the programs, i already >can directly find fixes for them. Simply not knowing cache line sizes >and such... Some programmer may not know this. But certainly this is not a top secret too. I respect all others. Even though they may not know everything, but certainly they know something ( that you and me don't know). btw, i have no idea what kind of hashtable you are using. But i know the popular one in chess is not the best one, at least for me, for both my chess program and my xiangqi program. Ren.
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.