Author: Vine Smith
Date: 15:30:54 04/28/01
Go up one level in this thread
On April 28, 2001 at 09:41:47, Uri Blass wrote: >On April 28, 2001 at 09:01:21, Joshua Lee wrote: > >> >>>>1.Nf3 Nf6 2.d4 d5 3.Nc3 e6 4.Bg5 Bb4 5.Qd3 Be7 >>>> = (0.00) Depth: 21/51 01:28:08 2030981kN >>>>1.d4! >>>> = (0.03) Depth: 21/51 01:36:20 2236387kN >>>>1.d4 Nf6 2.Nf3 d5 3.Nc3 e6 4.Qd3 Bb4 5.a3 Bxc3+ >>>> = (0.06) Depth: 22/58 05:46:18 8061585kN >>>>1.d4 Nf6 2.Nf3 d5 3.Nc3 e6 4.Qd3 c5 5.e4 cxd4 >>>> = (0.00) Depth: 23/56 09:41:19 13723728kN >>>>1.d4 Nf6 2.Nf3 d5 3.Nc3 e6 4.Qd3 c5 5.e4 Nc6 >>>> = (0.00) Depth: 24/59 16:03:57 23200850kN >>>> >>>>(Lee, 27.04.2001) >>>No, this isn't any book opening, even an obscure one. I've often thought it >>>would be interesting to compile a list of deep searches at move 1 from various >>>programs. But it's so disappointing -- 12 full moves deep, and this is the best >>>that it sees? Playing 3.Nc3 really saps all the tactical potential out of >>>White's position, and 4.Qd3 is just horrible. I guess it doesn't lose material, >>>since Fritz has looked quite deep, but long term, White will have problems. By >>>the way, did you mention what hardware this is, where you can hit ply 24? >> >> >>The computer i am using is a Compaq Presario Laptop with a PIII 500 and 192Mb of >>Ram with 128Mb for Hash Tables. Geting to this depth wasn't a problem the >>problem is that now it says 25/25 and it's at 42Hours i am going to try to give >>it as long as i can untill i get tired but it should start this ply by 48Hours. >>If not it's just too high to get much further without a faster system possibly >>someone with more than 2Ghz to spare and quite a bit of time could try the same >>thing. 512Ghz would be nice. I wondered how deep any of the top programs would >>need to play the opening without the book well as far as the latest fritz engine >>goes it's somewhere over 24ply. > >I am not sure. > >1.d4 is not a mistake and is one of the good moves so I do not see why do you >assume that 24 plies are not enough to play well. > >I suspect also that 0.00 is a correct score(the only possible correct scores are >0.00 or mate in some moves for one side). > >The opening position is relative simple and in more complicated positions 24 >plies may take weeks or even monthes or years to calculate(I know that there are >positions from my correspondence game when Deep Fritz or tiger need more than 24 >hours to get a score at depth 18. > >Uri From the perspective of selecting a good first move, Fritz has done fine. But look at what its choice is based on! The analysis is close to pure drivel, although when I checked at the chesslab.com online database, I did find 3 instances of the position after White's 4th move (.5 - 2.5 in Black's favor). I think the problem is that most engines are developed assuming that there will be a book, and they are not equipped to deal with the "artificial" starting position. So Fritz's premature attempt to take over the center is really just due to this being the proper way to deal with totally unbooked openings (say, if Black had played 1...a6 and 2...c6), whereas all Black attempts to contest the center (whether classically or hypermodern style) are assumed to be in book. I believe there's one Winboard engine, Tao, that plays without a book at first (it builds its own based on experience), but even here, it seems difficult to program good opening play (one game where it had White pathetically started 1.d4 Nf6 2.Nf3 e6 3.Bg5 h6 4.Bd2?). I think the dream of programs finding good opening lines from the start position may be as unrealistic as the notion of 10-man tablebases addressed elsewhere in this forum. Many opening lines are justified by variations that go too deep for any computer at any speed to ever search through. So I don't believe that Junior, or any "positional" type of program will produce better results. But I think I'll give some engines a crack at this kind of search anyway, just to see if anything interesting turns up. Perhaps we could classify them in this manner, e.g. 1.e4 at 20-ply engines vs. 1.d4 at 20-ply engines, as if we were assigning family and species groupings to animals.
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.