Author: Ed Schröder
Date: 03:17:46 01/26/00
Go up one level in this thread
On January 26, 2000 at 01:38:07, Christophe Theron wrote: >On January 26, 2000 at 00:42:20, Robert Hyatt wrote: > >>On January 25, 2000 at 21:54:44, Christophe Theron wrote: >> >>>On January 25, 2000 at 14:40:57, Jeremiah Penery wrote: >>> >>>>On January 25, 2000 at 14:02:01, Christophe Theron wrote: >>>> >>>>>Actually many programs prove everyday that using an imperfect pruning scheme is >>>>>much better than no pruning at all. Yours included (I don't mean your pruning is >>>>>imperfect). >>>>> >>>>>I see no reason to believe that this would change just because you are be able >>>>>to compute 4, 5 or even 10 plies deeper. >>>>> >>>>>Is it really still debatable? It was maybe, 20 years ago, but now we all know >>>>>the answer... >>>> >>>><snip> >>>> >>>>>My main point is that not using a pruning scheme is somewhat... stup... >>>>>Ahem... not really optimal. >>>> >>>><snip> >>>> >>>>>But I'm sorry, not using any pruning scheme, if it is really what they did, >>>>>sounds like a "political" decision. I cannot believe that Hsu is stupid enough >>>>>to really believe that DB plays better without pruning. >>>>> >>>>>It is either a huge professional mistake or a deliberate public relations >>>>>choice. You guess. >>>> >>>><snip> >>>> >>>>>OK, he had all this power as a result of his very hard work. So he deserved it. >>>>>But the total package could have been much better with a pruning scheme. The >>>>>thing that has played against Kasparov was far from being finished. >>>> >>>>Even the best selective-search algorithm will sometimes cause you to play a >>>>different move than you would without it. You will 'miss something', so to >>>>speak. If you were writing a program to play against arguably the best chess >>>>player _ever_, would you want to risk this happening even once and losing >>>>because of it? >>> >>> >>>Of course I would implement a selective search. >>> >>>It has been the main source of improvements in computer chess for the last 2 >>>decades. >>> >>>The drawbacks are very small when compared to the advantages. >>> >>>This is clearly demonstrated by the success of microcomputer chess programs >>>since Psion chess in 1981. And even before. >> >>maybe or maybe not. In 1970 we were _all_ selective searchers. Chess 4.x >>changed that in 1975. By 1980 nobody that was winning was selective. Then >>we went the other way with null-move and so forth. Maybe the right way is >>not risky forward pruning, when you have enough horsepower. Which sounds >>_exactly_ like what Slate/Atkin said in 1975 when you think about it. :) > > >You are right, if you have enough computational power to search to the end of >the game, you don't need any pruning system. > >But this is not the case here. And as far as I know, computing deeper, even if >imperfectly, increases your program's strength. > >Most of the programmers here, and most of the users of chess programs, would >certainly have a word or two to say to Slate and Atkin on this subject. :) > >Anyway the case "you have enough horsepower" is not very interesting. If you >mean by this that if you can beat your opponent without pruning you don't need >to prune, then you definitely have a point. That's not "maybe ot maybe not", >that's for sure this time. > > > > >>>> Especially if you had the speed not to really worry about it, as >>>>they did. Also, it may have been a matter of time. Since there was little time >>>>to do much testing, it would've been way too hard to write and debug a good >>>>selective search algorithm. They had other problems to worry about. >>> >>> >>>The lack of time is indeed the reason I think. >> >>I think it was lack of interest. Hsu was determined to do it without any >>inaccurate forward pruning. We could second-guess all we want. But don't >>lose sight of the fact that what he did worked well enough to beat Kasparov. > > > >I don't forget this. > >I'm just saying that the thing could have been much stronger if they had worked >on implementing a good pruning scheme rather than this SE thing. The log-files IMO clearly proof DB isn't a brute force program. The branch factor is much to low for that. Period. IMO. Ed > > >>>>In effect, they did do a bit of selective search, through their extensions. You >>>>could call their search a highly-selective 30-ply+ search. :) They just were >>>>not selective in the first part of the search. >>> >>> >>>It is a related idea, but it's not the same. >> >> >>The effect is essentially the same. Either extend some branches far deeper >>than others, or cut some off far earlier than others. Both tend to make the >>tree taller but narrower, which is a 'human-like' goal that may or may not be >>good over the long-haul. > >>>>I do agree that they may have been a bit better if they had some sort of >>>>selective search, however. I just don't think it was practical or really >>>>important for them to worry about. >>> >>> >>>Because they had no serious competition. Kasparov could have punished them, but >>>he failed too, unfortunately. Not that I think he is not able to. But on this >>>particular occasion, he failed. >>> >>> >>> >>> Christophe >> >> >>I don't think he had a prayer of punishing them tactically. Perhaps >>positionally, but then that is unrelated to search. And I think he would >>punish any other program tactically _and_ positionally. IMHO of course. > > >Maybe or maybe not. > >My interest was rather in the fact that they used a probably suboptimal approach >in a multimillion dollars project. > >Not sure anymore if you are interested in talking about this... > > > > Christophe
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.