Author: Vincent Diepeveen
Date: 16:34:12 12/24/02
Go up one level in this thread
On December 24, 2002 at 19:21:29, Dana Turnmire wrote: without wanting to sound rude, but aren't you confusing the word 'positional' and 'strategical' a bit here and there? if you make a strategical testset, then it measures something strategical obviously. If you make a strategical testset, then it doesn't measure in first instance something positional obviously. No one will stop you making a testset. They are always appreciated. It is hard work though. Most IMs are doing things pretty quick. preferably positions must get taken where just 1 move is best or where 1 move is complete idiocy to play. That may contain many easy positions for mankind too. Some will be amazingly hard for computers! I personally have not that much time to bugfix others testsets, so a testset which is very well analyzed before getting used is preferred bigtime. I was looking at some kind of CCC (I,II,III) testset this afternoon and missed to see for example why some Nxa6 move was much better than b4 move there which DIEP preferred also after a long think. Not knowing who checked out that CCC (I,II,III) testset i feel then that i can only run the testset at DIEP when i invest a lot of time checking out whether Nxa6 is better or b4 is better. Where better doesn't mean 'win in 20 versus win in 40'. A win is a win. 3 minutes a position is an excellent time to test a program. Keep doing the good job! >I am beginning to think there is no way to test programs positionally because no >matter what position is given it seems there is some objection to it (apples >compared to oranges, no one correct move, computer gave the right move for the >wrong reason etc.) > >Most agree that a strong human player understands long term strategy whereas >computers are dumb in this area. This is just one example out of the book >entitled "Multiple Choice Chess" by IM Graeme Buckley. When it comes to tactics >even the world champion would get blown away by the present day programs but >when it comes to strategy humans rule. Would this format be a satisfactory way >to test programs for making good positional moves? All the programs I used were >tested at 3min/move. > > >[Date "1995.01.20"] >[White "Ward, Christopher"] >[Black "Voskanian, Vahagn"] >[Result "1-0"] > >1.d4 Nf6 2.c4 e6 3.Nc3 d5 4.cxd5 exd5 5.Bg5 Be7 6.e3 Nbd7 7.Bd3 Nf8 >8.Qc2 Ne6 9.Bh4 c6 10.Nf3 g6 11.O-O O-O 12.Rab1 Ng7 13.b4 a6 14.a4 Bf5 >15.Ne5 Bxd3 16.Nxd3 Nf5 17.Bxf6 Bxf6 18.b5 axb5 19.axb5 Nd6 20.bxc6 bxc6 >21.Nb4 Qd7 22.Na4 Nc4 23.Nc5 Qd6 24.Rb3 Rfb8 25.Nba6 Rb5 26.Rxb5 cxb5 >27.Nb4 Bd8 28.g3 Bb6 29.Nbd3 Bxc5 30.dxc5 Qf6 31.Nb4 Rc8 32.c6 d4 >33.exd4 Qxd4 34.Rd1 Qe5 35.Qd3 Qe6 36.Qd5 Ne5 37.Rc1 Ng4 38.Qd4 Qe2 >39.c7 h5 40.Nd5 1-0 > >Stage one is complete. Black has a weak backward pawn on a semi-open file. >21.Nd1 and 21.Rfc1 are safe and both score one point. 21.Nb4 begins the attack >on black's weak link and bags two points. The weakest of White's choices was >21.f3, which scores no points. Take a bonus point if you spotted Black's strong >reply 21...Nc4 when the e3-pawn proves impossible to defend satisfactorily. For >example 22.Qc1 Re8 23.Re1 Nxe3!! 24.Rxe3 Bxd4 25.Nd1 Qe7 26.Kf2 Qh4+ 27.Kg1 Rxe3 >28.Nxe3 Re8. Take two bonus points if you saw 23...Nxe3 for Black. > >Arasan 6.1 - 21.Rfc1 >CM9000 - 21.Rb4 >Genius 7 - 21.Ra1 >Ruffian 21.Nc5 >[d]r2q1rk1/5p1p/2pn1bp1/3p4/3P4/2NNP3/2Q2PPP/1R3RK1 w - - 0 21 > >22.e4 only scores one point; it just offers mass exchanges with 22...Nxe4 23Nxe4 >dxe4 24.Qxe4 Qxd4 25.Qxd4 Bxd4 26.Nxc6 when a draw is a likely result. > >22.Rfd1 and 22.Ne2 are also sufficient for one point but 22.Na4 deserves two >points in view of the forking threat on b6 and the additional attack on c6. > >Arasan 6.1 - 22.Nd3 >*CM9000 - 22.Na4 >*Genius 7 - 22.Na4 >Ruffian - 22.Nd3 >[d]r4rk1/3q1p1p/2pn1bp1/3p4/1N1P4/2N1P3/2Q2PPP/1R3RK1 w - - 0 22 > >White still has to watch ...Na3 by Black and therefore must avoid 24.Rfc1, which >loses two points. > >The question is where should the rooks go? There is a useful possibility that >White can double after 24.Rb3 (two points) so this move slightly outscores the >other two which are worth one point. > >Arasan 6.1 - 24.Ne4 >CM9000 - 24.Ne4 >Genius 7 - 24.Ne4 >Ruffian - 24.Ne4 >[d]r4rk1/5p1p/2pq1bp1/2Np4/1NnP4/4P3/2Q2PPP/1R3RK1 w - - 0 24 > >After Rfb1 Na3 White finds he has fallen for the fork (this loses two points). >25.Nca6 and 25.Re1 avoid this (both score one) but with the former White has >made no progress after 25.Nca6 Rb6 26.Nc5. > >The best move must be 25.Nba6 (two points); this knight to a6 threatens to take >control of the b-file and ...Rb6 is obviously prevented. > >Arasan 6.1 - 25.Rc1 >CM9000 - Rc1 >*Genius 7 - 25.Nba6 >Ruffian - 25.Rc1 >[d]rr4k1/5p1p/2pq1bp1/2Np4/1NnP4/1R2P3/2Q2PPP/5RK1 w - - 0 25 > >It does not make sense to give away control of an open file, so 26.Rc3 does not >score. 26.Rfb1 is fine for one point. This beats 26.Rb4, which allows >26...Rxa6 27.Nxa6 Rxb4 28.Nxb4 Qxb4 (lose one point). > >The best, but only just, is 26.Rxb5 which splits the black pawns after the >recapture. In this way White exchanges one advantage (Black's backward pawn) >for another (the split b- and d-pawns). In addition, Black's d-pawn will become >isloated. Score two point for this. > >*Arasan 6.1 - 26.Rxb5 >CM9000 - 26.Rfb1 >*Genius 7 - 26.Rxb5 >Ruffian - 26.Rfb1 >[d]r5k1/5p1p/N1pq1bp1/1rNp4/2nP4/1R2P3/2Q2PPP/5RK1 w - - 0 26 > >27.Qb3 and 27.Rb1 allow the respective forks...Nd2 and ...Na3 and lose one and >two points respectively. 27.Ra1 merits one point, but leaves the knight on a6 >rather needlessly in a pin. 27.Nb4 places the knight securely in front of >Black's passed pawn (two points). > >*Arasan 6.1 - 27.Nb4 >*CM9000 - 27.Nb4 >Genius 7 - 27.Ra1 (Genius chooses 27.Nb4 at 4' 44") >Ruffian - 27.Ra1 >[d]r5k1/5p1p/N2q1bp1/1pNp4/2nP4/4P3/2Q2PPP/5RK1 w - - 0 27 > >28.h3 is okay for one point, but another pawn move is required if Black plays >...Bc7. 28.g3 carefully avoids the possible future checkmate threat and is >worth two points. 28.Qe2 does not have a logical plan behind it but it is safe >so it scores one point anyway. The same cannot be said for 28.Nb7; after >28...Qxb4 White finds that he has blundered a piece and three points. > >Arasan 6.1 - 28.Rc1 >CM9000 - 28.Nbd3 >*Genius 7 - 28.g3 >Ruffian - 28.Rb1 >[d]r2b2k1/5p1p/3q2p1/1pNp4/1NnP4/4P3/2Q2PPP/5RK1 w - - 0 28 > >Black has just captured so it would take something of a severe loss of >concentration to choose either 30.Kg2 or 30.Nf4 both of which lose three points. > 30.Nxc5 scores one point and leaves Black with the aforementioned pawn >structure, which cannot be bad for White. 30.dxc5 scores two points because it >gives White a passed pawn into the bargain. However, the nature of the game >changes slightly after this move. If the passed pawn is not defended, then it >is conceivable that it could be rounded up and taken. > >*Arasan 6.1 - 30.dxc5 >Genius 7 - 30.Nxc5 >*CM9000 - 30.dxc5 >*Ruffian - 30.dxc5 >[d]r5k1/5p1p/3q2p1/1pbp4/2nP4/3NP1P1/2Q2P1P/5RK1 w - - 0 30 > >31.Nb4 scores two points as it is an impressive multi-purpose move. White now >defends the square c6 to aid in the advance of the c-pawn. Black has a passed >pawn too, but for the moment its path is blocked and thus the b-pawn is not >looking dangerous. One final point of 31.Nb4 is that it also attacks the >d5-pawn. 31.Nf4 is an inferior version of this and scores one point. > >31.c6 is a straightforward blunder and loses one point. More complicated is >31.Qb3 which loses two points after 31...Nd2. If you saw further with 32.Qxd5 >Rd8! when Black cashes in more than the exchange, then give yourself one bonus >point. > >Arasan 6.1 - 31.Rd1 >*CM9000 - 31.Nb4 >*Genius 7 - 31.Nb4 >*Ruffian - 31.Nb4 >[d]r5k1/5p1p/5qp1/1pPp4/2n5/3NP1P1/2Q2P1P/5RK1 w - - 0 31 > >The idea is to centralise the queen and help the passed pawn up the board. >Playing 35.Qd3, without spotting 35...Nb2, reduces your score by two. You had >to see ahead and intend to play 36.Qd4 (or Qd5, Qd6); these three moves score >three bonus points. Best of all, 36.Qd7 socres four bonus points. > >35.Qb3 is safe and earns one point. 35.Qe2 loses nine points after 35...Qxe2 >which, by comparison, makes 35.Nd3 a good move, yet after 35...Qe4 it still >carries a penalty of one point. > >Arasan 6.1 - 35.Qd3 >CM9000 - 35.Qd3 >Genius 7 - 35.Qd3 >Ruffian - 35.Qd3 >[d]2r3k1/5p1p/2P3p1/1p2q3/1Nn5/6P1/2Q2P1P/3R2K1 w - - 0 35
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.