Author: Uri Blass
Date: 20:08:29 08/02/01
Go up one level in this thread
On August 02, 2001 at 22:01:08, Robert Hyatt wrote:
>On August 02, 2001 at 13:29:49, Uri Blass wrote:
>
>>On August 02, 2001 at 10:23:35, Robert Hyatt wrote:
>>
>>>On August 02, 2001 at 02:04:15, Hristo wrote:
>>>
>>>>On August 01, 2001 at 23:39:29, Robert Hyatt wrote:
>>>>
>>>>>On August 01, 2001 at 17:29:54, Andrew Williams wrote:
>>>>>
>>>>>>The game went:
>>>>>>
>>>>>>[Event "ICS Rated standard match"]
>>>>>>[Site "204.178.125.65"]
>>>>>>[Date "2001.08.01"]
>>>>>>[Round "-"]
>>>>>>[White "tartaruga"]
>>>>>>[Black "Dreev"]
>>>>>>[Result "0-1"]
>>>>>>[WhiteElo "2568"]
>>>>>>[BlackElo "2835"]
>>>>>>[TimeControl "7200+10"]
>>>>>>
>>>>>>1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 Nc6 6. Bc4 e6 7. Be3 Be7
>>>>>>8. f4 O-O 9. Qf3 Nxd4 10. Bxd4 Qa5 11. O-O e5 12. Be3 exf4 13. Qxf4 Be6 14.
>>>>>>Nd5 Bxd5 15. exd5 Rae8 16. b4 Qc7 17. Bd3 Nd7 18. Bd4 Ne5 19. Qg3 g6 20.
>>>>>>Bb5 Rc8 21. c3 f5 22. a4 Kh8 23. a5 a6 24. Be2 Bf6 25. Qf4 Qg7 26. Rad1 Nf7
>>>>>>27. b5 axb5 28. Bxb5 Be5 29. Qd2 Rc7 30. Rde1 Rfc8 31. Re2 Qf6 32. Qb2 Kg7
>>>>>>33. Bxe5 Nxe5 34. c4 Re7 35. Kh1 Rcc7 36. Rfe1 f4 37. Re4 g5 38. Qd4 Rc5
>>>>>>39. Qf2 Rc8 40. Qd2 Rcc7 41. Qe2 Ng6 42. Rf1 Rxe4 43. Qxe4 Re7 44. Qf3 Qb2
>>>>>>45. Rg1 Qd2 46. Qg4 Kf6 47. h3 Qxa5 48. Qc8 Re1 49. Kh2 Rxg1 50. Qe6+ Kg7
>>>>>>51. Kxg1 Qa1+ 52. Kh2 Qe5 53. Qd7+ Kh6 54. Kh1 Qe1+ 55. Kh2 Qg3+ 56. Kh1 f3
>>>>>>57. gxf3 Qe1+ 58. Kh2 Qf2+ 59. Kh1 Qxf3+ 60. Kh2 Qe2+ 61. Kg1 Qe1+ 62. Kh2
>>>>>>Qe2+ 63. Kg1 Nf4 64. Qg4 Qe1+ 65. Kh2 Qf2+ 66. Kh1 b6 67. h4 Qxh4+ 68.
>>>>>>Qxh4+ gxh4 69. Kg1 Kg5 70. Kf2 Kf5 71. Kf3 Ke5 72. Kg4 h3 73. Kg3 Kd4 74.
>>>>>>Kh2 Nd3 75. Kxh3 Ne5 76. Kh4 Nxc4
>>>>>>{tartaruga resigns} 0-1
>>>>>>
>>>>>>
>>>>>>Deep Fritz was on a Dual 1GHz PIII
>>>>>>
>>>>>>Andrew
>>>>>
>>>>>
>>>>>Maybe Frans will get off that marketing hype crap and stop with the
>>>>>"deep fritz is as good or better than Deep Blue and is ready for Kramnik"
>>>>>junk now.
>>>>
>>>>Why should he stop? He probably doesn't care that it is _only_ hype, so long as,
>>>>he makes money at the end! Why should his _hype_ be considered more _junk_ than
>>>>IBM's hype?
>>>>
>>>>cheers Bob. ;-)
>>>>hristo
>>>
>>>
>>>My problem with his statement is that I _know_ that he knows it is false, if
>>>he really made it in the first place. He _knows_ because he has played them
>>>more than once OTB at ACM/WCCC events.
>>
>>The last time that he played against them he used only Fritz3 on p90.
>>Knowing that Deep thought is better than Fritz3(p90) does not mean knowing that
>>Deeper blue is better than Deep Fritz on good SMP machine.
>
>Just do the math. He played DT more than once at 2M nodes per second.
I know about only one game that Fritz3(p90) played against Deep thought and
old hardware and software do not convince me.
DB was
>100x faster than DT in NPS, and another factor of 5-10X faster due to the
>additional evaluation they did at no cost in speed. Say a factor of 500-1000
>times faster. Do you think DF has _really_ made any headway at reducing that
>gap? IE what machine today is 100X faster than a P90? Much less 500X or
>1000X? There are none that are intel-compatible.
You consider the evaluation as being 5-10 times faster and you forget that the
evaluation of Deep Fritz is also clearly better than Fritz3.
>
>
>>
>>I agree that Deeper blue was better than Deep thought but I do not know if it
>>was practically the same as being 100 times faster.
>
>It is just the same as crafty at 1M vs crafty at 100M. Both were parallel. But
>DB had more evaluation.
>
>
>
>>It is possible that it was practically only at the same level of being 10 times
>>faster than Deep thought because of bugs.
>
>It is possible it is 1M times faster due to bugs too. I discount that as fritz
>can _also_ have bugs (it almost certainly does, as I still find them in Crafty
>and I have done this longer than Frans).
>
>
>
>>
>>I am not sure if Deep Fritz with 2M nodes per second is going to lose in a match
>>against old programs like Fritz3 with 200M nodes per second at tournament time
>>control.
>
>fritz 3 has _nothing_ to do with deep blue 2. DB 2 was created in 1997. That
>is only 4 years old. It certainly was not nearly as primitive at fritz at that
>time...
I do not know if it's evaluation was better than Fritz3's evaluation.
Hsu did not release Deeper blue's evaluation.
You explained a case when Deep thought understood that an opposite color bishop
endgame is won.
It does not prove that Deeper blue had better evaluation because it is possible
that Deeper blue's evaluation was better in the endgame but not in the middle
game.
>
>
>
>
>
>>
>>One reason is better search rules of Deep fritz and another reason is better
>>evaluation of Deep Fritz.
>
>
>You think DF has a better evaluation that DB? Based on what? I had the
>opportunity to talk with them many times. I know of the significant volume
>of knowledge they had built in.
Volume is not quality and I know at least one position when Deep Fritz has
better evaluation(Deeper blue could not see Qe3 in it's pv when Deep Fritz has
no problem to see it).
I mentioned the bishops (opposite) + pawns
>knowledge that was clearly of grandmaster quality since it came from a GM.
>I don't know of _any_ program that does that currently besides them. There
>are many other such things about DB.
>
>
>
>
>>
>>I believe that Deep Fritz has also better evaluation and better search rules
>>than Deeper blue.
>>
>
>Depends on whether you believe in null-move or not. I do. Hsu didn't. Ed
>apparently doesn't. So what? It obviously worked. I haven't seen any micro
>win game six vs black against a strong human. I doubt any micro could have
>drawn at least two of those difficult endings against Kasparov. I don't see
>a single thing that suggests that DF has a better evaluation. I see it make
>really silly computer-type mistakes with pawn structure and endgames. I didn't
>see DB make those nearly as often, if at all.
Deeper blue did not play a lot of games so people could not learn about Deeper
blue's weaknesses.
Deep Fritz played a lot of games and it is the difference.
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.