Author: Robert Hyatt
Date: 15:02:10 11/15/02
Go up one level in this thread
On November 15, 2002 at 16:26:22, Uri Blass wrote: >On November 15, 2002 at 14:56:26, Rolf Tueschen wrote: > >>On November 15, 2002 at 10:36:54, Uri Blass wrote: >> >>>On November 15, 2002 at 10:26:03, Robert Hyatt wrote: >>> >>>>On November 15, 2002 at 07:39:00, Uri Blass wrote: >>>> >>>>>On November 15, 2002 at 06:30:17, Rolf Tueschen wrote: >>>>> >>>>>>Uri, >>>>>>I hope that you will agree if we make a deeper analysis of the whole complex, >>>>>>are you game? Thanks, and don't take anything of that too personal. It isn't. >>>>>>The truth is that I am reflecting such questions since long, and from a science >>>>>>view I see questions over questions but few answers. >>>>>> >>>>>>November 15, 2002 at 01:33:16, Uri Blass wrote: >>>>>> >>>>>>>The reason that I say that I believe that the programs of today are better is >>>>>>>some analysis of the games with the logfiles. >>>>>> >>>>>>Ok, that's a good idea. However - >>>>>> >>>>>> >>>>>>> >>>>>>>The claim that the programs of today are better is not a claim that is only a >>>>>>>claim of some programmers. >>>>>> >>>>>>Who said that? Not programmers but PR spin doctors. Totally different world. >>>>>> >>>>>>> >>>>>>>There are logfiles of the games and in a few cases it is possible to compare >>>>>>>times that deeper blue needs to see something with time that Deep Fritz need to >>>>>>>see the same thing. >>>>>> >>>>>>You are a bit talking around in approximate terms. What logfiles and where. Why >>>>>>only in a few cases. That sounds synthetic. Usually you gave much better >>>>>>evidence. >>>>> >>>>>I give here some data. >>>>> >>>>>The problem is that in most of the cases there are not similiar lines but in >>>>>cases that there are similiar lines or similiar fail low it is possible to >>>>>compare which program does it faster. >>>>> >>>>>In order to find more data you need to analyze every position in the logfile >>>>>when only in part of the cases you may find similiar lines or similiar behaviour >>>>>because the programs have different evaluation. >>>>> >>>>>You can download the logfiles of the games >>>>> >>>>>Here are the logfiles of game 1 >>>>> >>>>>http://www.research.ibm.com/deepblue/watch/html/game1clean.log >>>>> >>>>>Case 1: >>>>>Deep blue pondered about 40.gxh5 that was not played by kasparov >>>>>Deep blue failed low again and again >>>>> >>>>>-96 T=65 Bf3h5p rf1h1 Kh6g5 rh1g1 Kg5h6 bc3e5N Re8e5b bh3g4 Bh5g6p pf5g6B Kh6g6p >>>>>bg4d7 Kg6f7 bd7c6P 11(6) #[Bxh5](-116)############################## -116 T=229 >>>>>Bf3h5p rf1h1 Kh6g5 rh1g1 Kg5h6 pg6g7 Ne5f3 rg1g3 Pc6c5 pb4c5P Rd8c8 pf5f6 Rc8c5p >>>>>bh3d7 12(6) #[Bxh5](-146)[Bxh5](-146) -146v T=311 Bf3h5p rf1h1 Kh6g5 rh1g1 Kg5h6 >>>>>pg6g7 Ne5f3 rg1g3 Pc6c5 pb4c5P Rd8c8 pf5f6 Rc8c5p pg7g8/q Re8g8q rg3g8R Rc5c3b >>>>>rg8h8 12(6) #[Bxh5](-171)# 'rg1' --------------------------------------- --> >>>>>rf1g1 <-- --------------------------------------- -171 T=538 Bf3h5p rf1h1 Kh6g5 >>>>>rh1g1 Kg5h6 pg6g7 Ne5f3 rg1g3 Nf3g5 bc3f6 Rd8d2 kf2e1 Rd2e2 ke1f1 Ng5h3b rg3h3N >>>>>Re2e3n rh3e3R >>>>> >>>>> >>>>>Deep Fritz also fails low again and again after a similiar time with similiar >>>>>score and it gives the feeling that the programs are of similiar strength. >>>>> >>>>> >>>>>Kasparov,G - Deep Blue >>>>>3rr3/8/2p3Pk/1p2nP1P/pP2p3/P1B1Nb1B/2P2K2/5R2 b - - 0 1 >>>>> >>>>>Analysis by Deep Fritz: >>>>> >>>>>40...Ne5xg6-- >>>>> ² (0.59) Depth: 1/3 00:00:00 >>>>>40...Ne5xg6-- 41.h5xg6 >>>>> +- (3.56) Depth: 1/4 00:00:00 >>>>>40...Kh6xh5! >>>>> +- (3.34) Depth: 1/4 00:00:00 >>>>>40...Kh6xh5! >>>>> ² (0.41) Depth: 1/4 00:00:00 >>>>>40...Kh6xh5 41.Rf1-g1 >>>>> ² (0.50) Depth: 2/8 00:00:00 >>>>>40...Kh6xh5 41.Rf1-g1 Kh5-h6 >>>>> ² (0.50) Depth: 3/11 00:00:00 >>>>>40...Kh6xh5 41.Rf1-g1 Kh5-h6 42.g6-g7 >>>>> ² (0.62) Depth: 4/11 00:00:00 1kN >>>>>40...Bf3xh5! >>>>> ² (0.59) Depth: 4/13 00:00:00 2kN >>>>>40...Bf3xh5! 41.Bh3-g2 Ne5xg6 42.f5xg6 Kh6xg6 >>>>> ² (0.34) Depth: 4/16 00:00:00 3kN >>>>>40...Bf3xh5 41.Rf1-g1 Ne5xg6 42.f5xg6 Bh5xg6 >>>>> ² (0.44) Depth: 5/14 00:00:00 7kN >>>>>40...Bf3xh5 41.Rf1-g1 Ne5-f3 42.Rg1-g3 Re8-g8 43.Bh3-g2 >>>>> ² (0.56) Depth: 6/15 00:00:00 14kN >>>>>40...Bf3xh5 41.Rf1-g1 Ne5-f3 42.Rg1-g3 Re8-g8 43.Bh3-g2 Rd8-f8 >>>>> ² (0.62) Depth: 7/19 00:00:00 43kN >>>>>40...Bf3xh5 41.Rf1-g1 Ne5-f3 42.Rg1-g3 Re8-g8 43.g6-g7 Kh6-h7 44.f5-f6 >>>>> ² (0.69) Depth: 8/21 00:00:00 88kN >>>>>40...Bf3xh5 41.Rf1-h1 Kh6-g5 42.Rh1-g1+ Kg5-h6 43.g6-g7 Ne5-f3 44.Rg1-g3 Re8-g8 >>>>> ± (0.72) Depth: 9/24 00:00:00 595kN >>>>>40...Bf3xh5-- >>>>> ± (1.03) Depth: 10/27 00:00:01 995kN >>>>>40...Bf3xh5-- >>>>> ± (1.03) Depth: 10/29 00:00:02 1317kN >>>>>40...Bf3xh5 41.Rf1-h1 Kh6-g5 42.Rh1-g1+ Kg5-h6 43.g6-g7 Ne5-f3 44.Ne3-g4+ Bh5xg4 >>>>>45.Rg1xg4 Kh6-h7 46.Kf2-e3 >>>>> ± (1.06) Depth: 11/28 00:00:07 4515kN >>>>>40...Bf3xh5 41.Rf1-h1 Kh6-g5 42.Rh1-g1+ Kg5-h6 43.g6-g7 Ne5-f3 44.Ne3-g4+ Bh5xg4 >>>>>45.Rg1xg4 Kh6-h7 46.Kf2-e3 >>>>> ± (1.06) Depth: 12/29 00:00:12 8140kN >>>>>40...Bf3xh5 41.Kf2-g3 Kh6-g5 42.g6-g7 Ne5-f3 43.f5-f6 Bh5-f7 44.Bh3-g4 Nf3-d2 >>>>>45.Rf1-f5+ Kg5-h6 46.Bc3xd2 >>>>> ± (1.25) Depth: 13/32 00:00:31 20418kN >>>>>40...Bf3xh5 41.Kf2-g3 Kh6-g5 42.g6-g7 Ne5-f3 43.f5-f6 Bh5-f7 44.Bh3-g4 Nf3-d2 >>>>>45.Rf1-f5+ Kg5-h6 46.Rf5-c5 >>>>> ± (1.31) Depth: 14/33 00:01:12 47451kN >>>>>40...Bf3xh5-- >>>>> +- (1.62) Depth: 15/37 00:03:08 122687kN >>>>>40...Bf3xh5-- >>>>> +- (1.62) Depth: 15/40 00:04:28 175879kN >>>>>40...Bf3xh5 41.Rf1-h1 Kh6-g5 42.Rh1-g1+ Kg5-h6 43.g6-g7 Kh6-h7 44.f5-f6 Ne5-f7 >>>>>45.Ne3-f5 Rd8-d1 46.Rg1xd1 >>>>> +- (1.72) Depth: 16/39 00:08:42 345199kN >>>>> >>>>>(Blass, Tel-aviv 15.11.2002) >>>>> >>>>> >>>>>case 2: >>>>> >>>>>Both Deep Fritz(1000Mhz 64 Mbytes hash) and Deeper blue change their main line >>>>>from 42...Nxg4+ 43.Rxg4 Kh5 to 42...Nxg4+ 43.Rxg4 Rd5 at similiar time Deep >>>>>fritz is sligthly faster. >>>>> >>>>> >>>>>Kasparov,G - Deep Blue >>>>>3rr3/8/2p3Pk/1p2nP2/pP2p1N1/P1B5/2P2K2/6R1 b - - 0 1 >>>>> >>>>>Analysis by Deep Fritz: >>>>> >>>>> >>>>> >>>>>42...Ne5xg4+ 43.Rg1xg4 Kh6-h5 44.Rg4-g1 Rd8-d5 45.f5-f6 Rd5-f5+ 46.Kf2-e1 Re8-g8 >>>>>47.f6-f7 Rg8xg6 48.Rg1-f1 >>>>> +- (1.53) Depth: 15/34 00:00:45 29095kN >>>>>42...Ne5xg4+ 43.Rg1xg4 Rd8-d5 44.f5-f6 Rd5-f5+ 45.Kf2-e2 Re8-g8 46.g6-g7 Kh6-h5 >>>>>47.Rg4xe4 Kh5-g6 48.Re4-e6 >>>>> +- (1.50) Depth: 16/36 00:01:56 74709kN >>>>> >>>>>(Blass, Tel-aviv 15.11.2002) >>>>> >>>>> >>>>>T=51 Ne5g4n rg1g4N Kh6h5 rg4g1 Re8f8 pf5f6 Rf8g8 pg6g7 Rd8d5 kf2e3 Rd5d7 bc3e5 >>>>>10(6) #[Nxg4](-129)[Nxg4](-129) -129v T=77 Ne5g4n rg1g4N Kh6h5 rg4g1 Re8f8 pf5f6 >>>>>Rf8g8 pg6g7 Rd8d5 kf2e3 Rd5f5 ke3e4P 10(6) #[Nxg4](-151)### -151 T=201 Ne5g4n >>>>>rg1g4N Rd8d5 pf5f6 Rd5f5 kf2e2 Re8g8 pg6g7 Kh6h5 rg4g2 Kh5h6 ke2e3 Rf5f1 ke3e4P >>>>>Kh6h7 rg2h2 11(6) 'ng4' [cont] -151 T=222 Ne5g4n rg1g4N Rd8d5 pf5f6 Rd5f5 kf2e2 >>>>>Re8g8 pg6g7 Kh6h5 rg4g2 Rf5f3 bc3b2 Kh5h6 rg2g4 Rf3f5 bb2d4 Pe4e3 ke2e3P Kh6h5 >>>>>rg4f4 --------------------------------------- >>>>> >>>>> >>>>>case 3: >>>>> >>>>>Analysis of the end of game 1 >>>>> >>>>> >>>>>I find near the end of game 1 >>>>> >>>>>T=73 Rd5f5 kf2e2 Re8g8 pg6g7 Kh6h5 rg4g1 Rf5f3 bc3b2 Kh5h6 rg1g4 Rf3f5 bb2d4 >>>>>Pe4e3 ke2e3P Rf5f1 ke3e4 Rf1f6p 11(6) >>>>> >>>>>#[Rf5](-260)v[find a move]#########[TIMEOUT][et3 1295 >>>>>sec]#[Rg8](-183)[Rd1](-180) -180 T=204 Rd5d1 pf6f7 Rd1f1 kf2f1R Kh6h5 pf7e8R/q >>>>>Kh5g4r --------------------------------------- --> 44. .. Rd1 >>>>> >>>>> >>>>>It is clear that deeper blue failed low after 204 seconds and a bug caused it to >>>>>play a random move but the fact that it failed low is because of Ke3 that does >>>>>not appear. >>>>> >>>>>Deep Fritz also shows similiar fail low >>>>> >>>>>Here is some analysis on 1000Mhz >>>>>It failed low after similiar time >>>>> >>>>>Kasparov,G - Deep Blue >>>>>4r3/8/2p2PPk/1p1r4/pP2p1R1/P1B5/2P2K2/8 b - - 0 1 >>>>> >>>>>Analysis by Deep Fritz: >>>>> >>>>>44...e4-e3+-- >>>>> = (0.19) Depth: 1/3 00:00:00 >>>>>44...e4-e3+-- 45.Kf2-e2 >>>>> ± (1.22) Depth: 1/5 00:00:00 >>>>>44...Rd5-d1! >>>>> ± (1.00) Depth: 1/5 00:00:00 >>>>>44...Rd5-d1-- >>>>> ± (1.37) Depth: 2/6 00:00:00 >>>>>44...Rd5-d1-- >>>>> ± (1.37) Depth: 2/6 00:00:00 >>>>>44...Re8-d8! >>>>> ± (1.28) Depth: 2/7 00:00:00 >>>>>44...Re8-g8! >>>>> ± (1.25) Depth: 2/8 00:00:00 >>>>>44...Re8-g8! 45.g6-g7 >>>>> ² (0.47) Depth: 2/11 00:00:00 >>>>>44...Re8-g8 45.g6-g7 Rd5-f5+ 46.Kf2-e2 >>>>> ± (0.72) Depth: 3/14 00:00:00 >>>>>44...Re8-g8 45.g6-g7 Rd5-f5+ 46.Kf2-e2 Rf5-f3 >>>>> ± (0.75) Depth: 4/12 00:00:00 2kN >>>>>44...Re8-g8 45.g6-g7 Rd5-f5+ 46.Kf2-e2 Rf5-f3 47.Bc3-d2+ Kh6-h7 >>>>> ± (0.75) Depth: 5/13 00:00:00 6kN >>>>>44...Re8-g8-- >>>>> ± (1.06) Depth: 6/16 00:00:00 14kN >>>>>44...Re8-g8-- 45.g6-g7 Rd5-h5 46.Rg4xe4 Rh5-h2+ 47.Kf2-e3 Rh2xc2 >>>>> +- (1.53) Depth: 6/17 00:00:00 20kN >>>>>44...Rd5-f5+! >>>>> +- (1.50) Depth: 6/18 00:00:00 36kN >>>>>44...e4-e3+! >>>>> +- (1.47) Depth: 6/18 00:00:00 44kN >>>>>44...e4-e3+! 45.Kf2-e2 Rd5-h5 46.g6-g7 Rh5-h2+ 47.Ke2-d3 Re8-d8+ 48.Bc3-d4 e3-e2 >>>>> ² (0.66) Depth: 6/18 00:00:00 51kN >>>>>44...e4-e3+-- >>>>> ± (0.97) Depth: 7/19 00:00:00 65kN >>>>>44...e4-e3+-- 45.Kf2-e2 Rd5-h5 46.g6-g7 Rh5-h2+ 47.Ke2-d3 Re8-d8+ 48.Rg4-d4 >>>>>Rd8xd4+ 49.Kd3xd4 Kh6-h7 50.Kd4xe3 >>>>> +- (1.47) Depth: 7/24 00:00:00 85kN >>>>>44...Rd5-d8! >>>>> +- (1.44) Depth: 7/24 00:00:00 146kN >>>>>44...Rd5-d8! 45.f6-f7 Re8-f8 46.Kf2-e3 Kh6-h5 47.Rg4-g3 Rd8-d6 >>>>> ± (1.25) Depth: 7/24 00:00:00 155kN >>>>>44...Rd5-d8-- >>>>> +- (1.56) Depth: 8/15 00:00:00 194kN >>>>>44...Rd5-d8-- 45.f6-f7 Re8-f8 46.Kf2-e3 Kh6-h5 47.Rg4-g2 Kh5-h6 48.Ke3xe4 >>>>> +- (2.56) Depth: 8/18 00:00:00 212kN >>>>>44...e4-e3+! >>>>> +- (2.53) Depth: 8/21 00:00:00 250kN >>>>>44...e4-e3+! 45.Kf2-e2 Rd5-f5 46.f6-f7 Rf5-f2+ 47.Ke2-d3 Re8-d8+ 48.Kd3xe3 >>>>>Rf2xc2 >>>>> +- (2.03) Depth: 8/23 00:00:00 307kN >>>>>44...Rd5-d7! >>>>> +- (2.00) Depth: 8/23 00:00:00 494kN >>>>>44...Rd5-d7! 45.g6-g7 e4-e3+ 46.Kf2-f3 Kh6-h7 47.Rg4-h4+ Kh7-g6 48.Rh4-e4 Re8-g8 >>>>>49.Kf3xe3 >>>>> +- (1.94) Depth: 8/23 00:00:00 515kN >>>>>44...Re8-d8! >>>>> +- (1.91) Depth: 8/23 00:00:00 553kN >>>>>44...Re8-d8! 45.g6-g7 Kh6-h7 46.Rg4xe4 Rd5-h5 47.Kf2-g3 Kh7-g8 48.Re4-e7 >>>>> +- (1.87) Depth: 8/23 00:00:00 572kN >>>>>44...Re8-g8! >>>>> +- (1.84) Depth: 8/23 00:00:01 642kN >>>>>44...Kh6-h5! >>>>> +- (1.81) Depth: 8/23 00:00:01 661kN >>>>>44...Kh6-h5-- >>>>> +- (2.12) Depth: 9/22 00:00:01 776kN >>>>>44...Kh6-h5-- 45.f6-f7 Re8-f8 46.Kf2-g3 Rd5-d8 47.Bc3-g7 e4-e3 >>>>> +- (2.50) Depth: 9/22 00:00:01 855kN >>>>>44...Re8-g8! >>>>> +- (2.47) Depth: 9/22 00:00:01 888kN >>>>>44...Re8-g8! 45.g6-g7 Rd5-f5+ 46.Kf2-e2 Kh6-h5 47.Rg4xe4 Kh5-g6 48.Re4-e6 c6-c5 >>>>>49.b4xc5 Rf5xc5 >>>>> +- (1.91) Depth: 9/22 00:00:01 971kN >>>>>44...Re8-g8 45.g6-g7 Rd5-f5+ 46.Kf2-e2 Kh6-h5 47.Rg4xe4 Kh5-g6 48.Re4-e6 c6-c5 >>>>>49.b4xc5 Rf5xc5 50.Ke2-d3 >>>>> +- (2.00) Depth: 10/23 00:00:02 1568kN >>>>>44...Kh6-h5! >>>>> +- (1.97) Depth: 10/27 00:00:03 2293kN >>>>>44...Kh6-h5! 45.f6-f7 Re8-f8 46.Kf2-g3 Rd5-d8 47.Bc3-g7 Rf8-g8 48.f7xg8Q >>>>> +- (1.81) Depth: 10/27 00:00:03 2362kN >>>>>44...Kh6-h5 45.f6-f7 Re8-f8 46.g6-g7 Rf8xf7+ 47.Kf2-e2 Rd5-d8 48.g7-g8Q Rd8xg8 >>>>>49.Rg4xg8 Rf7-f3 50.Bc3-b2 >>>>> +- (1.87) Depth: 11/25 00:00:04 2791kN >>>>>44...Kh6-h5-- >>>>> +- (2.19) Depth: 12/29 00:00:06 4330kN >>>>>44...Kh6-h5-- >>>>> +- (2.19) Depth: 12/29 00:00:06 4582kN >>>>>44...Re8-g8! >>>>> +- (2.16) Depth: 12/29 00:00:07 5091kN >>>>>44...Re8-g8! 45.g6-g7 Rd5-f5+ 46.Kf2-e3 Rf5-f3+ 47.Ke3-e2 Kh6-h7 48.Bc3-b2 >>>>>Rf3-f5 49.Ke2-e3 Rf5-f1 50.Rg4xe4 >>>>> +- (2.03) Depth: 12/29 00:00:09 6238kN >>>>>44...Rd5-f5+! >>>>> +- (2.00) Depth: 12/31 00:00:15 10217kN >>>>>44...Rd5-f5+! 45.Kf2-e2 Re8-g8 46.g6-g7 Kh6-h5 47.Rg4xe4 Kh5-g6 48.Re4-e6 c6-c5 >>>>>49.Re6-e4 c5xb4 50.Re4xb4 >>>>> +- (1.50) Depth: 12/31 00:00:16 10831kN >>>>>44...Rd5-f5+ 45.Kf2-e2 Re8-g8 46.g6-g7 Kh6-h5 47.Rg4xe4 Kh5-g6 48.Re4-e6 c6-c5 >>>>>49.Re6-e4 c5xb4 50.Re4xb4 >>>>> +- (1.50) Depth: 13/29 00:00:19 12786kN >>>>>44...Rd5-f5+-- >>>>> +- (1.81) Depth: 14/31 00:00:27 18278kN >>>>>44...Rd5-f5+-- 45.Kf2-e2 Re8-g8 46.g6-g7 Kh6-h5 47.Rg4xe4 Kh5-g6 48.Re4-e6 c6-c5 >>>>>49.b4xc5 Rf5xc5 50.Ke2-d3 >>>>> +- (1.84) Depth: 14/35 00:00:45 31295kN >>>>>44...Rd5-f5+ 45.Kf2-e2 Re8-g8 46.g6-g7 Kh6-h5 47.Rg4xe4 Kh5-g6 48.Re4-e6 c6-c5 >>>>>49.b4xc5 Rf5xc5 50.Ke2-d3 >>>>> +- (1.94) Depth: 15/33 00:01:21 55884kN >>>>>44...Rd5-f5+-- >>>>> +- (2.25) Depth: 16/37 00:03:29 141379kN >>>>>44...Rd5-f5+-- 45.Kf2-e3 Rf5-f3+ 46.Ke3-e2 Re8-g8 47.Bc3-d2+ Kh6-h5 48.Rg4-g5+ >>>>>Kh5-h6 49.f6-f7 Rg8-d8 50.g6-g7 >>>>> +- (3.06) Depth: 16/37 00:05:42 229369kN >>>>>44...Re8-g8! >>>>> +- (3.03) Depth: 16/37 00:06:04 242531kN >>>>> >>>>>(Blass, Tel-aviv 15.11.2002) >>>>> >>>>>Uri >>>> >>>> >>>>That is the "tactical side". However I have personally seen enough >>>>positional mistakes by deep fritz to not compare them at all in that >>>>area. DF is perfectly ok in _some_ positions. But in others, forget >>>>about it. _particularly_ in pawn endings where things like majorities >>>>have to be taken into account. And it simply doesn't... >>> >>>Pawn ending are not something that happen often in games and more knowledge >>>about pawn endgames may give Fritz probably not more than 10 elo. >>> >>>The most important part is the middle game and the main problem is that humans >>>do not know to define the right evaluation. >>> >>>Knowledge is not quantity of evaluation but quality and >>>years of experience give advantage to Fritz relative to the deeper blue team. >>> >>>The deeper blue team had the potential to produce better evaluation with more >>>time but I do not believe that their evaluation in 1997 was superior relative to >>>the programs of today. >>> >>>The possibility to put everything you want is an advantage but lack of time to >>>tune the evaluation is a disadvantage. >>> >>>Uri >> >>I was waiting for Bob to respond and stayed in silence. But Bob gave exactly the >>argument I expected, and you failed to reply! What you are saying is again >>another speculative reflection. But Bob took you to task with your beautifully >>selected positions. What is with the positions where FRITZ is weak? > >I see no reason to assume that there are more positions that Fritz is weak than >position that Deeper blue was weak. > >Deep Fritz can play Kh1 in game 2 when Deeper blue played Kf1. >Deep Fritz can see Qe3 in the main line when Deeper blue could not. That means _nothing_. Does fritz see the draw? Nope. A simple piece/square evaluation will like Qe3 as it gets the queen as close to the enemy king as possible. Whether it is a good move or not doesn't matter. Serendipity doesn't impress me. Any time someone posts a position here, a program will find an impossibly-hard-to-find move by serendipity, just because its eval happens to like the right move for the wrong reasons. That doesn't mean that program is better than the rest, however... I think there are too many moves that DB found that DF _won't_ find that would make the difference... > >There are clearly positions when Deep Fritz is better. Again, so what? If it is better in 2, and worse in 10, is it better or worse overall? > >Hyatt claim that Fritz was lucky and it does not understand the position but >good evaluation is exactly evaluation that make you lucky more times so the fact >tha Fritz plays Kh1 and not Kf1 in game 2 and the fact that Fritz can see Qe3 in >it's main line is an evidence for superiority of deep fritz(it is not a proof >but it is an evidence) Again, serendipity. Can fritz see that Kh1 leads to a draw? Not a chance. In that light, Kf1 is certainly a superior move to Kh1 for the obvious reasons. So liking Kh1 for the wrong reason is not exactly an endorsment of the program... Seeing Qe3 is trivial for a piece/square program or one that simply says "get queen close to king." Seeing Kh1 is bad if the program doesn't understand why as h1 is the worst square on the board for the king, in general... > >> >>Let me enforce that. I couldn't believe my eyes that you gave so many long >>chapters about a couple of positions. But why you have chosen exactly these and >>how _representative_ "these" are, is unknown, I think to you too. > >I did not analyze all the games to compare deep fritz with deeper blue but >I got the impression that deep blue was not better. > >> >>But ok, if you really want to prove your claim then it's time enough. So that is >>not a general "No!". > >I do not say that I can prove it but I can have an opinion based on data that I >see and my opinion is that deeper blue was not better than Deep Fritz. >> >>But I found it very interesting that Bob necessarily came to my own arguments >>finally. A program cannot be proven strong just with 5 or maybe 50 selected >>positions where it works. What is with those positions where the programs are >>weak??? At least against humans such weaker links have fatal consequences. A >>chain is only as strong as its weakest link. (Rolf Tueschen 1997; very old >>wisdom) > >I did not say that deep fritz has no position when it is weak but the same is >for deeper blue. > >Deeper blue could know things that Deep Fritz does not know about pawn endgames >but Deep Fritz may know other important things that deeper blue does not know(I >do not know what deeper blue knew about pawn endgames because it did not play >them in the match against kasparov so I can compare only based on the games and >I do not find impressive knowledge of Deep blue that deep fritz does not have >based on the games) > >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.