Author: Uri Blass
Date: 09:11:19 08/02/02
Go up one level in this thread
On August 02, 2002 at 11:02:24, Slater Wold wrote: >On August 02, 2002 at 09:41:51, Uri Blass wrote: > >>On August 02, 2002 at 07:54:53, Slater Wold wrote: >> >>>On August 02, 2002 at 07:50:31, pavel wrote: >>> >>>>On August 02, 2002 at 05:47:52, K. Burcham wrote: >>>> >>>>> >>>>> >>>>>Rxc4 >>>>> >>>>>In any case, it was a breathtaking move and I still marvel at it more than any >>>>>other move I have seen a chess program play. >>>>>Dan Corbit >>>>> >>>>>Dan do we know the Rxc4 line in Junior eval? >>>>>I could not get Deep Junior7 to play this move. 2x1533mhz, 12 hours. >>>>>Why are you so fascinated by this move when we have not proven this >>>>>move is a good sacrifice? have you seen eval change in favour of black with >>>>>another program? i have not seen this. >>>>>kburcham >>>> >>>>It was a differant program than Junior7. >>>>According to Amir, Junior8 will most likely be released at the end of this year >>>>(during Fall), and it will be aprox. 50 elo points better than it's previous >>>>version. >>>> >>>> >>>>So it's probably is not a good idea to see if Junior7 finds Junior8's move. >>>> >>>>By the way, can anyone be kind enough to provide the position (again) in this >>>>thread, I wanna give it another shot by other programs, to see if there was a >>>>better move. >>>> >>>>To K.burcham: do you suggest/know any other better/best move than Rxc4? >>>> >>>>cheers, >>>>pavs >>> >>>[D]r1r5/3nbpk1/4qnp1/4p1Bp/1pN1P3/1PP2Q2/4N1PP/3R1R1K b - - >>> >>>I have followed 2 lines down to a win for Junior. Rd5 and h3. >>> >>>In about 10 hours of "evaluating", I would say Rxc4 is the best move. >>> >>>It is just incredibly deep. >> >>Can you prove that Rxc4 is best by yace and a tree? >>Yace is the only program that I know that has a good learning function >>during analysis. >> >>if you start by analyzing the leaves of the tree and go backwards yace can agree >>that a move is best inspite of not finding it when it gets only the root >>position. >> >>There are other programs that can learn during analysis but usually their >>learning is inferior and if the tree has more than one line and I start by going >>backwards in line A and continue by going backward in line B they may forget >>line A because they use the rule:first forget what you learned first(it is >>impossible for programs to remember everything in analysis because the tree is >>large and programs must forget part of the information). >> >>Yace knows that score,main line and depth is something that it should never >>forget in analysis and it can use the information to have better analysis of the >>root position. >> >>Uri > >I didn't try it with YACE. But I did do backwards (and forwards) analysis with >Shredder, Fritz, Junior, Crafty, and Rebel. And they all agreed, after Nxc4 xc4 >b3 the chances of winning or drawing are slim to none. Of course, not all of >them see that so clearly, at that particular point. > >But perhaps when I get home from work, I will give Yace a try. I suggest that you first write the tree that you analyze and start from the leaves of the tree and go with yace slowly backward from the leaves to the root. The problem here is that after Rxc4 bxc4 b3 white has a lot of options and not only Rd5 and h3 that are the options that you mentioned. A good tree to prove that black is winning should include at least one move after every logical reply of white for b3 and programs considers a lot of options of white as good and not only Rd5 or h3 that you mentioned. I suggest that before giving yace a tree you first find all the candidate moves for white after Rxc4 bxc4 b3 and put a reply to them in the tree. The best way to do it is to start with Deep fritz and give it one hour to analyze. If the socre is not at least 0.5 pawns for black click next best and give it again an hour to analyze. If the score is still not at least 0.5 pawns for black do it again until you get 0.5 pawns for black. It does not seem to be an easy task because some short analysis tells me that at depth 13 there are more than 9 moves with a positive score based on Deep Fritz's analysis. I hope that it is possible to reduce the number significantly if you use fast hardware and next best but I am not sure. [D]r7/3nbpk1/4qnp1/4p1Bp/2P1P3/1pP2Q2/4N1PP/3R1R1K w - - 0 1 Here is analysis by Deep Fritz with 9 options at depth 11,12,13 and all the moves have positive score. Analysis by Deep Fritz(depth 11): 1. ± (0.72): 3.Rb1 Qxc4 4.Ng3 Ra2 5.Rbd1 Qe6 6.c4 Nc5 7.Rb1 2. ² (0.66): 3.Rd5 Nb6 4.Nc1 b2 5.Nd3 Nxc4 6.Nc5 Qc6 7.Rdd1 Rb8 8.Rb1 Bxc5 3. ² (0.63): 3.Rd2 Qg4 4.Qe3 Kg8 5.Bxf6 Nxf6 6.Nc1 Nxe4 7.Re2 Nc5 8.Qxe5 4. ² (0.63): 3.h3 Ra4 4.Ng3 Nc5 5.Kg1 Rxc4 6.Rb1 Qb6 7.Rf2 5. ² (0.59): 3.Nf4 exf4 4.e5 Ng4 5.Bxe7 Re8 6.Bf6+ Ndxf6 7.exf6+ Qxf6 8.Rd4 Ne5 6. ² (0.59): 3.Nc1 Qxc4 4.Ne2 Qe6 5.c4 b2 6.Rd2 Ra3 7.Rd3 7. ² (0.59): 3.c5 Ra6 4.Bxf6+ Nxf6 5.Nc1 Qc4 6.Rde1 Bxc5 7.Nd3 Nd7 8. ² (0.56): 3.Ng3 Qg4 4.Qe3 h4 5.Ne2 Ra2 6.Rf3 Kf8 7.Rdf1 Bc5 9. ² (0.53): 3.Rde1 (Blass, Tel-aviv 02.08.2002) r7/3nbpk1/4qnp1/4p1Bp/2P1P3/1pP2Q2/4N1PP/3R1R1K w - - 0 1 Analysis by Deep Fritz(depth 12): 1. ± (0.72): 3.Rb1 Qxc4 4.Ng3 Ra2 5.Rbd1 Qe6 6.c4 Nc5 7.Rb1 Ng4 2. ² (0.66): 3.c5 Nxc5 4.Nd4 exd4 5.e5 Ng4 6.Bxe7 Qxe7 7.cxd4 3. ² (0.63): 3.h3 Ra4 4.Ng3 Nc5 5.Rb1 Rxc4 6.Rb2 Ra4 4. ² (0.59): 3.Rd5 Nb6 4.Rb1 Nbxd5 5.cxd5 Qb6 6.Bxf6+ Bxf6 7.c4 Rb8 8.Nc1 5. ² (0.59): 3.Rd2 Qg4 4.Qe3 Kg8 5.Bxf6 Nxf6 6.Nc1 Qxe4 7.Qb6 6. ² (0.53): 3.Nc1 Qxc4 4.Nxb3 Qxb3 5.Rxd7 Nxd7 6.Bxe7 Qe6 7.Bg5 Ra2 7. ² (0.44): 3.Ng3 Qg4 4.Qxg4 hxg4 5.Rd2 Nc5 6.Bxf6+ Bxf6 7.Rd5 Ra5 8.Kg1 8. ² (0.41): 3.Nf4 exf4 4.e5 Ng4 5.Bxe7 Re8 6.Bf6+ Ndxf6 7.exf6+ Qxf6 8.c5 Ne3 9. ² (0.41): 3.Rde1 Ra4 4.h3 Rxc4 5.Ng3 Qc6 6.Re3 Qe6 7.Rd1 Nc5 8.Kg1 (Blass, Tel-aviv 02.08.2002) r7/3nbpk1/4qnp1/4p1Bp/2P1P3/1pP2Q2/4N1PP/3R1R1K w - - 0 1 Analysis by Deep Fritz(depth 13) iteration was not finished: 1. ± (0.75): 3.Ng3 Ra6 4.Rd5 Ba3 5.Bxf6+ Nxf6 6.Rb5 b2 7.Qd3 Rd6 2. ² (0.56): 3.Rb1 Qxc4 4.Ng3 Ra2 5.Rbd1 Qe6 6.c4 Nc5 7.Rb1 3. ² (0.50): 3.h3 Ra4 4.Ng3 Nc5 5.Ra1 Rxc4 6.Ra7 Ng8 4. ² (0.50): 3.Nc1 Qxc4 4.Nxb3 Qxb3 5.Rxd7 Nxd7 6.Bxe7 Qe6 7.Bg5 f6 8.Be3 Ra3 5. ² (0.44): 3.Rd5 Nb6 4.Rb1 Nbxd5 5.cxd5 Qb6 6.Bxf6+ Bxf6 7.c4 Rb8 8.Nc1 b2 6. ² (0.44): 3.Rde1 Ra6 4.h3 Nc5 5.Ng3 Rb6 6.Ra1 Ng8 7.Be3 7. ² (0.41): 3.Rd2 Qg4 4.Qe3 Qe6 5.Rdd1 8. ² (0.38): 3.Nf4 exf4 4.e5 Ng4 5.Bxe7 Ra2 6.Bd6 Ne3 7.Qxf4 Nxd1 9. = (0.22): 3.c5 Nxc5 4.Nd4 exd4 5.e5 Ng4 6.Qxa8 dxc3 7.Bxe7 (Blass, Tel-aviv 02.08.2002) Ng1 and Rfe1 are found as better than c5 in that iteration but Ng1 is not good enough to be included in the best 9. r7/3nbpk1/4qnp1/4p1Bp/2P1P3/1pP2Q2/4N1PP/3R1R1K w - - 0 1(still iteration not finished) Analysis by Deep Fritz: 1. ± (0.75): 3.Ng3 Ra6 4.Rd5 Ba3 5.Bxf6+ Nxf6 6.Rb5 b2 7.Qd3 Rd6 2. ² (0.56): 3.Rb1 Qxc4 4.Ng3 Ra2 5.Rbd1 Qe6 6.c4 Nc5 7.Rb1 3. ² (0.50): 3.h3 Ra4 4.Ng3 Nc5 5.Ra1 Rxc4 6.Ra7 Ng8 4. ² (0.50): 3.Nc1 Qxc4 4.Nxb3 Qxb3 5.Rxd7 Nxd7 6.Bxe7 Qe6 7.Bg5 f6 8.Be3 Ra3 5. ² (0.44): 3.Rd5 Nb6 4.Rb1 Nbxd5 5.cxd5 Qb6 6.Bxf6+ Bxf6 7.c4 Rb8 8.Nc1 b2 6. ² (0.44): 3.Rde1 Ra6 4.h3 Nc5 5.Ng3 Rb6 6.Ra1 Ng8 7.Be3 7. ² (0.41): 3.Rd2 Qg4 4.Qe3 Qe6 5.Rdd1 8. ² (0.38): 3.Nf4 exf4 4.e5 Ng4 5.Bxe7 Ra2 6.Bd6 Ne3 7.Qxf4 Nxd1 9. ² (0.38): 3.Rfe1 Qg4 4.Qe3 Ra2 5.Ra1 Bc5 6.Bxf6+ Nxf6 7.Qxc5 Qxe2 (Blass, Tel-aviv 02.08.2002) Rfe1 is weaker than Bh4 Yace 0.99.56 - Blass,U r7/3nbpk1/4qnp1/4p1Bp/2P1P3/1pP2Q2/4N1PP/3R1R1K w - - 0 1 Analysis by Deep Fritz(end of iteration 13): 1. ± (0.75): 3.Ng3 Ra6 4.Rd5 Ba3 5.Bxf6+ Nxf6 6.Rb5 b2 7.Qd3 Rd6 2. ² (0.56): 3.Rb1 Qxc4 4.Ng3 Ra2 5.Rbd1 Qe6 6.c4 Nc5 7.Rb1 3. ² (0.50): 3.h3 Ra4 4.Ng3 Nc5 5.Ra1 Rxc4 6.Ra7 Ng8 4. ² (0.50): 3.Nc1 Qxc4 4.Nxb3 Qxb3 5.Rxd7 Nxd7 6.Bxe7 Qe6 7.Bg5 f6 8.Be3 Ra3 5. ² (0.44): 3.Rd5 Nb6 4.Rb1 Nbxd5 5.cxd5 Qb6 6.Bxf6+ Bxf6 7.c4 Rb8 8.Nc1 b2 6. ² (0.44): 3.Rde1 Ra6 4.h3 Nc5 5.Ng3 Rb6 6.Ra1 Ng8 7.Be3 7. ² (0.44): 3.Bh4 Qg4 4.Bxf6+ Nxf6 8. ² (0.41): 3.Rd2 Qg4 4.Qe3 Qe6 5.Rdd1 9. ² (0.38): 3.Nf4 exf4 4.e5 Ng4 5.Bxe7 Ra2 6.Bd6 Ne3 7.Qxf4 Nxd1 (Blass, Tel-aviv 02.08.2002) 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.