Author: Christopher R. Dorr
Date: 08:36:32 01/25/99
Go up one level in this thread
On January 24, 1999 at 17:15:37, Melvin S. Schwartz wrote: > >On January 19, 1999 at 18:39:13, Fernando Villegas wrote: > >>On January 19, 1999 at 16:41:05, Melvin S. Schwartz wrote: >> >>>In the review of the Novag Diamond, it is stated by Mr. Kaufman that with the >>>"Selectivity" feature on the computer rating is about 100 points higher. I am >>>somewhat baffled at this because the "Brute Force" search mode would seem to be >>>superior since it examines the position more extensively. Unfortunately, in the >>>manual of my Mephisto atlanta, it doesn't say which is stronger only that Brute >>>Force search minimizes the risk of an oversight. My question is whether Mr. >>>Kaufman is correct in his assumption that Selective search is superior to Brute >>>Force search? Also, if I remember correctly, Novag claimed according to Mr. >>>Kaufman that Selective "On" was superior. Is there an answer from some authority >>>as to whether Selective or Brute Force results in a higher rating? >> >> >>Hi: >>I am not authority in this but just an old fart that has had too many computers >>along too many time and read too many about this to suspect with some ground >>that brute or selective approaches has his strenghts and his weakneses, but in >>the long run a good selective approach seems to be better as much is capable of >>deeper searches and so can avoid traps, positional mistakes etc in better way >>that the brute force. There are another reasons long to explain and beside I am >>not the man to do so. This one, the brute force appraoch, is incarcerated inside >>a narrow horizon due to the time it expends analyzing all in each branch of the >>tree, but then this very same fact let the program to be more eficient in >>tactical shots inside his narrowed horizon. Nevertheless, as selective approach >>is not suicidical, rarely they commit mistakes that entails a fast defeat, but >>on the contrary a brute force approach is prone to commit any mistake possible >>beyond his horizon. So, selective seems to be superior and facts show it is so. >>Demise of Spracklen team was due, people said, to his unbreakable attachment to >>brute force or better said to a variation of it called full width search. >>I hope I am not too much mistaken in all this >>Greetings from Chile >>fernando >>Fernando >Hello, Fernando! >I have contacted a technical advisor from Saitek in Hong Kong who has assured me >that Brute Force is superior to Selective Search. His explanation was that Brute >Force examines the position more extensively and therefore minimizes the risk of >an oversight. I know this conflicts with your opinion, but, the manual does >state that Brute Force minimizes the risk of an oversight that might occur in >Selective Search. Mr. Feldman has not responded to my query about his article. >And that is a shame. By the way, my wife is from Ecuador. I have been there >three times on vacation. Love the weather in Quito. Unless this 'selective search' is different from conventional definitions of 'selective search', selective search is *most definitely* stronger than brute force. regardless of what the tech guy said. The reason is as follows (somewhat simplified). Brute force eveluates *every position* is every branch of the search tree to a certain depth, dictated by time constraints. Lets say that (in time control x) that the brute force mode is able to examine every single position arising from the current position up to depth 8 ply. That mean that the computer sees *every* possible position that can arise in the next 4 full moves, and make is decision based on what it can force within that search depth. The selective search mode utilizes it's time differently. It may well search the first (say) 4 or 5 or 6 plies completely (ala brute force), but it will selectively search out in certain lines much deeper (say to 10 or 12 or 14 ply). If the position is tactically intersting (and therefore extended), this mode may well find a win of a pawn on ply 6/14 (i.e. 6 ply brute force, extended to 14 ply in certain lines). There are, of course, trade offs with each method. The selective search may miss a tactic in a line that it prunes off at ply 5 at ply 8 in that line that brute force sees. This isn't terribly likely, as selective search will follow a given line down further as long as it's tactically 'interesting'. Brute force's trade off comes because it sees *nothing* beyond a certain depth, because it has to spend it's time examing every position in it's given search depth, regardless of how ridiculous or unpromising it might be. The trade offs with brute force generally cost a program *far more* rating points than do the trade offs of selective search. I've owned dozens of programs and dedicated chess computers, and cannot remember any of them playing as well with brute force search as with selective search. Chris
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.