Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Test Positions / WAC.230: Refuting the refutation of a refutation
- - By Dadi Jonsson (Silver) [is] Date 2010-08-03 19:16
WAC.230 is among those chess positions that chess engines have always had problems with. It comes from the game Hans Kmoch - Aaron Nimzowitsch, Bad Niendorf (Germany), 1927.  Nimzowitch published the game with his annotations in Die Praxis Meines Systems (1929). WAC.230 refers to position 230 in Fred Reinfeld's Win at Chess where the position after White's 50th move is shown with the comment "Another brilliant Nimzovich breakthrough".

2b5/1r6/2kBp1p1/p2pP1P1/2pP4/1pP3K1/1R3P2/8 b - -

For many years, users tried their chess engines on this position and an occasional "success" was reported although most engines failed to "solve" it. Then in 2001(?) Alex Szabo gave a refutation, saying that there was nothing more than draw for Black. After 1...Rb4 2.cxb4 a4 3.b5+ Kxb5 4.Ba3 c3

2b5/8/4p1p1/1k1pP1P1/p2P4/Bpp3K1/1R3P2/8 w - - 0 1

he suggested 5.Re2! instead of Kmoch's 5.Rb1, which lead to a loss after 5...Kc4. Alex continued 5...Kc4 6.f4 Kxd4 7.f5.

2b5/8/4p1p1/3pPPP1/p2k4/Bpp3K1/4R3/8 b - - 0 1

Here he gave the continuation 7...exf5 and he is correct that it leads to a draw. However, the variations after 7...Kd3 are much more interesting and it is not so clear how White can keep the balance. This was actually the continuation chosen by Stockfish 1.8, when Dann Corbit ran it a few days ago on WAC.230. What is even more interesting is that Stockfish scores the position as a win for Black! In Dann's words: "In 16 seconds it simultaneously finds both the solution, along with Alex Szabo's brilliant refutation, and then finds a refutation to the refutation". After 15(?) years of testing, it seems that we have gone the full circle and Nimzowitsch's maneuver was actually winning. That would have been a nice end to this long story where several generations of chess engines, running on amazingly powerful hardware, have failed.

We have some brilliant analysts here on Rybka forum and a couple of them had a look at this position last year. Before revealing their conclusions, let's have a look at the variation produced by Stockfish. It agrees with the mainline presented above, and then it continues (after 7...Kd3):

8.fxg6 Kxe2 9.g7 b2 10.g8=Q b1=Q 11.Qxc8 Qg1+ 12.Kh4 Qh1+ 13.Kg3 Qf3+ 14.Kh4 Qe4+

2Q5/8/4p3/3pP1P1/p3q2K/B1p5/4k3/8 w - -

So far, so good, but now Stockfish fails with the losing 15.Kg3. Our analysts continued 15.Kh5 instead and then

15...Qh7+ 16.Kg4 d4
2Q5/7q/4p3/4P1P1/p2p2K1/B1p5/4k3/8 w - -

This is a critical position for the evaluation of the variation and if your engine failed at finding the solution starting from the initial position of WAC.230, perhaps you should try it a few moves into the main variation (even as late as here) and see how it does. According to the excellent analysis by Charles Sullivan and Mark Rawlings, posted here on the forum, the position is still drawn at this point, but the key move is 17.Kc6!!, drawing. Other moves lose. It looks like Charles and Mark have refuted Stockfish' refutation to the refutation. Actually, they "pre-refuted" the variation that Stockfish came up with since their analysis was posted a year ago :lol:

Charles (one of the best analysts posting here on the forum) posted his analysis here.

Mark used more or less automatic Rybka 3 based IDeA analysis, first starting from the initial position and then focusing on the position after 1...Rb4. His analysis is a treasure for those who are interested in WAC.230. Even if he set IDeA to analyze only to depth=15, it seems to me that all the questions you might have are answered in the resulting analysis tree. This includes the refutation of the Stockfish variation posted by Dann. Mark's IDeA analysis is a "required reading" for anyone wanting to analyze WAC.230. He was kind enough to post his analysis tree so anyone can download it.
Parent - By Uly (Gold) [mx] Date 2010-08-03 20:37

> Other moves lose. It looks like Charles and Mark have refuted Stockfish' refutation to the refutation. Actually, they "pre-refuted" the variation that Stockfish came up with since their analysis was posted a year ago :lol:


Ouch, my head! I suggest we call the "refutations of a refutation" a metarefutation :), and then they refute THAT...
Up Topic The Rybka Lounge / Test Positions / WAC.230: Refuting the refutation of a refutation

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill