Interesting CTB Error

When using the program Chess Tactics for Beginners, one shouldn't unquestioningly accept their answers. Sometimes there are alternate solutions to the problems that aren't accepted as correct answers. But you should also question their evaluations of variations.

The following problem was interesting. White to draw:



Note: for the following analysis, I'm trying to abide by the Nunn Convention for annotating endgames, hence the extensive use of "!". It's not just me being over-excited. (Technically, some of them might really be "!?", because other moves waste time but don't actually give away the win, but close enough.)

The move 1.Ng1! indeed does lead to a draw, as shown by 1...c1(Q) =. However, the software also gives the following line as a draw: 1...Kf5 2.Ne2 Ke4 3.Kg1 Kd3 4.Kf1 Kd2=. I've checked this with Fritz, and 1...Kf5? actually appears to lose for Black! The main theme appears to be that the white Knight can prevent the c-pawn from promoting (either from e2 or a2, as required) while White queens the h-pawn.

After 1.Ng1!= Kf5?-+ 2.Ne2! Ke4 3.Kg1! Kd3 4.Kf2! (instead of CTB's 4.Kf1?=):



For example, 4...Kd2 5.Kf3! Kd1 6.Nc3+:



It seems clear that Black cannot queen the c-pawn. However, I wasn't ready to trust Fritz yet. I wondered: is it possible that Black can draw the rook-pawn endgame, as I discussed in my previous post? Apparently not. One line continues: 6...Kd2 7.Na2 Ke1 (7...Kd3?! 8.Nb4+. This is a common theme in this, and other knight endgames: some squares such as d3 are "mined" because a king on that square falls victim to a knight fork that wins the offending pawn.) 8.Kg3 Ke2 9.Kxh3 Kf3 10.Kh4 Kf4 11.Kh5 Kf5 12.h4 Kf6 13.Kh6 Kf7 14.h5 Kg8 15.Kg6 Kh8 16.h6 Kg8:



We've followed the drawing procedure for K+RP vs. K, and it looks like Black will draw. However, after 17.Nc1! Kh8 18.Nd3! Kg8 19.h7! Kh8 20.Ne5!:



20...c1=Q 21.Nf7#.

Totally freakin' cool.