Author: Uri Blass
Date: 00:59:12 06/07/01
Go up one level in this thread
On June 07, 2001 at 01:15:43, Robert Hyatt wrote: >On June 06, 2001 at 15:29:59, Uri Blass wrote: > >>On June 06, 2001 at 11:13:28, Robert Hyatt wrote: >> >>>On June 06, 2001 at 07:26:28, Uri Blass wrote: >>> >>>>On June 06, 2001 at 06:58:28, Gian-Carlo Pascutto wrote: >>>> >>>>>...is that Fritz played the Dutch open just last year, on a quad, >>>>>and it didn't win, it didn't even got second. It got _third_, and >>>>>this while several players forfeited vs it in protest. >>>>> >>>>>We're only one year of development further now, and the hardware >>>>>will be twice as fast, *no more*. Not even a ply more for Fritzie. >>>>> >>>>>The Dutch are pretty ok as far as chess is concerned, but it's >>>>>not like they have any players that are remotely as strong as >>>>>Kramnik. >>>>> >>>>>It was not in the top 2 of Dutch players, and now it is supposed >>>>>to stand a chance vs. the World Champion? Uhh, well maybe if he >>>>>gets sick or crazy or something. >>>> >>>>It is clear that Deep Fritz has no chance against Kramnik. >>>> >>>>Nobody expects Fritz to win. >>>> >>>>I guess that most people are not going to bet that Fritz is going to beat >>>>Kramnik even if they can earn 100$ for winning and lose 1$ for losing. >>>> >>>>Fritz is public and it is unfair. >>>> >>>>A secret program when Kramnik know nothing about it may have chances but noy >>>>Fritz. >>>> >>>>Deeper blue was a secret program when kasparov knew nothing about it. >>>>Deep Fritz is not close to be a secret program. >>> >>>Kasparov didn't "know nothing" about it. DB Jr had played many exhibition >>>matches. >> >>DB Jr is not known to be something close to Deeper blue so I doubt if using Deep >>blue JR games could help kasparov. >> >>I also know that DB jr played games that were not tournament time control but >>something like active chess so I doubt if kasparov could learn much about deeper >>blue weaknesses. >> > > >DB Jr was identical to deep blue, just slower. Same eval, same everything. >Do you think that you couldn't learn anything about how to beat a computer >in a standard time control game by playing blitz games with it? I could >certainly find weaknesses... You can find weaknesses but the problem is if the weaknesses are the same at longer time control. when you do not know the evaluation of the machine you cannot know if the weaknesses are the same at tournament time control. You can see the mistakes of the program but you cannot see the evaluation of the program. Seeing blitz games with evaluation of the machine is different and you can learn more about the machine. Even in this case the evaluation may be misleading because it is possible that a program learn to change the evaluation function during the game(for example it may use tiger14's evaluation and when it can see that something is wrong based on the scores to try another evaluation). See the following example from my correspondence game against the female world champion in postal chess(we are at move 21 now and I believe that this analysis cannot help my opponent): Luba kristol - Uri blass [D]r2qkb1r/5pp1/p3p2p/4n3/1p1NP1P1/4B3/PPPQ3P/2KRR3 b kq - 0 1 Analysis by Chess Tiger 14.0: 17...Nxg4 18.Bg1 Qc7 19.Qe2 Ne5 20.Rf1 Bc5 µ (-1.40) Depth: 7 00:00:00 77kN 17...Nxg4 18.Bf4 e5 19.Rg1 h5 20.h3 Qxd4 21.Qxd4 exd4 22.hxg4 hxg4 23.Rxd4 µ (-1.30) Depth: 8 00:00:01 231kN 17...Nxg4 18.Bf4 e5 19.Rg1 h5 20.h3 Qxd4 21.Qxd4 exd4 22.hxg4 hxg4 23.Rxd4 µ (-1.30) Depth: 9 00:00:02 408kN 17...Nxg4 18.Bf4 e5 19.Rg1 h5 20.Nc6 Qc7 21.Bg5 Bc5 µ (-1.30) Depth: 10 00:00:03 737kN 17...Nxg4 18.Bf4 e5 19.Rg1 exf4 20.Rxg4 g5 21.Rg2 Bc5 22.Nf5 Qxd2+ 23.Rgxd2 a5 µ (-1.14) Depth: 11 00:00:08 1681kN 17...Nxg4 18.Bf4 e5 19.Rg1 exf4 20.Rxg4 g5 21.Rg2 Bc5 22.Nf5 Qxd2+ 23.Rgxd2 a5 24.Nd6+ Kf8 µ (-1.14) Depth: 12 00:00:22 4551kN 17...Nxg4 18.Bf4 e5 19.h3 exf4 20.hxg4 Bc5 21.Rf1 0-0 22.Rxf4 Qc7 23.e5 Qxe5 µ (-1.12) Depth: 13 00:00:46 9463kN 17...Nxg4 18.Bf4 e5 19.Qe2 exf4 20.Qxg4 Qg5 21.Qf3 Bc5 22.e5 Rc8 23.e6 0-0 24.exf7+ Rxf7 25.Ne6 µ (-1.18) Depth: 14 00:02:07 26849kN 17...Nxg4 18.Bf4 e5 19.h3 exf4 20.Qxf4 Nf6 21.Nc6 Qc8 22.Qe5+ Qe6 23.Qc7 Nd7 24.Qb7 Rd8 25.Nxd8 Kxd8 µ (-1.04) Depth: 15 00:04:19 55267kN We see that the evaluation of tiger goes down almost every iteration inspite of the fact that the score is positive so it seems that something is wrong with tiger's evaluation so it may be a good idea to pick another evaluation. Let pick gandalf's evaluation Analysis by Gandalf 4.32h: 17...Nxg4 18.Rg1 Nxe3 19.Qxe3 Qc7 20.Qe2 Qf4+ 21.Kb1 g6 = (0.20) Depth: 7 00:00:02 250kN 17...Bc5 = (0.19) Depth: 7 00:00:04 500kN 17...Bc5 18.Qg2 Qh4 19.h3 0-0 20.Rf1 a5 = (-0.04) Depth: 7 00:00:06 1100kN 17...Bc5 18.Bg1 Qc7 19.Qg2 0-0 20.g5 hxg5 21.Qxg5 a5 = (-0.14) Depth: 8 00:00:15 2900kN 17...Bc5 18.Qf2 Nxg4 19.Qg3 Nxe3 20.Qxg7 Nxd1 21.Qxh8+ Kd7 22.Qxd8+ = (-0.17) Depth: 9 00:00:55 11600kN 17...Bc5 18.Qf2 Qa5 19.Kb1 0-0 20.Rg1 Nc4 21.g5 e5 22.gxh6 ³ (-0.27) Depth: 10 00:02:10 28950kN 17...Bc5 18.Qe2 Qb6 19.Rd2 0-0-0 20.Red1 Rd7 21.h3 Rhd8 22.Qf2 = (-0.24) Depth: 11 00:06:15 83400kN you can see that Gandalf does not suffer from the same problem. 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.