George Duke wrote on Tue, May 7, 2013 07:24 PM EDT:
[Added 8.May: Okay '16 Nxe4...' leaves the other Rook to be captured, so it is fair to analyze for improvement from Black's Move 16 on; that is, White Move 20 is the second ''brilliancy'' not the first. Then 16 ...Nxa1 is the turnaround for Black to win. Gone is the a-Rook that gets moved three times in a row, and the passed d-Pawn is not organized to be threat. Black Queen stays mobile enough to keep the upper hand.]
Http://www.chessgames.com/perl/chessgame?gid=1284171.
Move '21 Nxe6...' by post-mortem parfait guidelines, dispensing or
avoiding the hackneyed opening theory phase (beat to death for five hundred
years), is the definitional brilliancy. The reason is that it dramatically
leaves the Rook in jeopardy with prospective exchange sacrifice. Somehow
the improvement that consistently is able to be justified to reverse the
outcome happens the first move or two afterwards and not later. So here in present 1938 Parr-Wheatcroft, how about '21 ...Nxg2' instead of actual '21 ...Nxf1'? That way taking Bishop removes important piece in the finale orchestrated by White, whilst the Rook may not have opportunity to figure.
It's not enough to suggest the improvement. Rather can 21 ...Nxg2 be justified for Black to win? Otherwise have to find a different improvement one of the Moves 21, 22, 23 actually made post technical not subjective brilliancy.