Author: blass uri
Date: 15:25:17 07/28/00
Go up one level in this thread
On July 28, 2000 at 17:49:14, Jeremiah Penery wrote: >On July 28, 2000 at 17:08:53, blass uri wrote: > >>On July 28, 2000 at 16:47:27, Jeremiah Penery wrote: >> >>>On July 28, 2000 at 16:04:43, Christophe Theron wrote: >>> >>>>No need to marvel at the singular extension stuff. When you compute 100 times >>>>faster than your opponent, you kill him anyway. But it's not sexy enough, and >>> >>>This isn't necessarily true when the opponent is human. DB was built to play >>>humans (Kasparov), and the team thought that the regular ideas would not be good >>>enough to beat him. They implemented singular (and other) extensions they >>>thought would help them compete against Kasparov. >>> >>>>Hsu and co needed something more flashy. Singular extensions does the trick: >>>>1) it is not done by anybody else (for the good reason that it is ineffective). >>> >>>Several others have used it. I know Bruce at least uses it now, and Lang has >>>used some form of singular extensions in his programs in the past. >>> >>>>2) we can afford to use it as we are so superior in NPS anyway. >>> >>>True, and also they thought it was the only way they could compete against >>>Kasparov. Because if Kasparov sees deeper than you in some important/forcing >>>line, he will probably beat you. >> >>I disagree because most of the tactics are not singular and kasparov will >>probably not win the commercial of today by seeing some plies deeper in singular >>lines. > >I think most tactics are at least partly singular. If Kasparov sees deeper than >you in tactics he will beat you. The DB team wanted to take no chances of >things like this happening. Deep Junior played against the best players(except kasparov) and the games prove that singular extensions is not important. > >>I am not impressed by their extensions because thir extensions could not help >>them to see simple tactics against Fritz3. > >That was an older version of their hardware and software. Let's not compare >this to DB '97. I can compare it to DB 97 because they used their singular extensions also in 95. The hardware was slower but it only proves that their extensions were not so good because they did not help them with 2M nodes per seconds. Or I can find some games of old Rebel/Junior/Shredder/... >versions where they played very bad blunders. > >>Their extensions also did not help them to see the draw against Kasparov in game >>2 and I expect extensions to help in this position. > >The draw is just way too deep for any computer to see. It is too deep for computers to see because they do not use the play against yourself extension. It was easy to see for a team of human and computer. I found it with the help of Genius3(p100) in 1997. The point is that I expected them to see more than the commercial programs by extensions. The fact that they failed to do it is an evidence that their extensions were not productive. I expect extensions to be productive in games and not in the nolot test and they missed the opportunity that could convince me that their extensions are productive. > >>The fact that they did not expect the drawing move of kasparov Qe3 when part of >>the top programs can see it suggests that their evaluation is also inferior. > >What is the evaluation of "the top programs" for Qe3 compared to a different >move? Fritz5.32 that is not the latest Fritz gives the following evaluation: Qxc6 -1.62 depth 14/30 fail high Qe3 -1.59 depth 14/41 solved fail high with -1.50 score. If you set up the position at move 46 with white to move, how long does >it take these programs to play a different move than Ra6 because they see Qe3? They cannot play different move than Ra6 but at least they expect Qe3 after enough time(they need less nodes than the number of nodes that deeper blue calculated without expecting the right move). 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.