There’s a few points I’d like to clear up. First, you don’t need a rating of 2100 to consider the possibility of 45 ... h5. If you look at my earlier posts you’ll see I considered it but (incorrectly) thought 46 Kf5 would draw.

Then there’s the issue of silicon assistance. The computer I’m using isn’t mine so I’m not at liberty to download software onto it. To me white’s position looks hopelessly lost after <44…h5 45.Kd4 Kd7 46.Kd5 c6+ 47.Kc5 Ke6 48.Kxc6 Kxe5> (for example, 49 Kd7 Kf4 50 Ke6 g5) so I’d appreciate a sample continuation to show how white saves it.

Finally, let me attempt a more helpful response to your question. What I have in my head is difficult to put into words, but in essence the answer is something along the lines that given the choice between a complex line (i.e. 44 ... h5) and a win that is more easily demonstrable (i.e. 44 ... Kd7) then the <general rule> is to pick the latter. Now this sounds patronising and I don’t mean it to be and it also relies on considering 44 ... Kd7 to be an easily demonstrable win, which is a bit subjective.

So what makes me claim that 44 ... Kd7 is an easily demonstrable win? When I chose it, it was on the basis that white is without a decent move. He can’t go after black’s K-side pawns yet (which 44 ... h5 allows), he can’t push home his own e-pawn and the remaining option leads to the game line which is not too difficult to analyse.

This is a plausible continuation of the text from move 45 on.

45 Kf5 c5 46 Kf4 c4 47 Ke4 c3 48 Kd3 Ke6 49 Kxc3 Kxc5 50 h5.

click for larger view

Per the Nalimov table bases, in this position there is only one move that wins for black, 50…Ke4.

click for larger view

50…Kf4 instead, draws.

click for larger view

49.Kd7 Kf4 50.Ke7 Kg3 51.Kf7 Kxg2 52.<Kg6> >

You are correct - this is drawn. The weird thing is that I remember seeing Kg6 when I took a look at it last night, but when I started to write out a possible solution in Notepad this morning, I forgot about it completely. So perhaps typing moves in Notepad does not confer such an unfair advantage after all.

<I would appreciate a helpful response to my previous question about a general principle from anyone to whom the puzzle is "obvious" or "not a puzzle at all".>

This discussion certainly establishes that it's possible for reasonably good players to go wrong in this ending and that accuracy is important at all times regardless of rating.

I think it's fair to say that this is technical position that should be won with accurate play, but what part of that makes this "not a puzzle"? CG is doing us a favor by throwing us an endgame once and a while, to enlighten those who suffer from endgame aversion.

Black will try and play for a draw. The most apparent goal is to get the c-file pawn free and block white's incoming e-file pawn.

Candidates: 44. ...Kd7

This is the only candidate that accomplishes the goal and doesn't put the king in check.

44. ...Kd7 it is.

Time to check.

To my mind a good puzzle needs to have just one solution. Today’s seems to have at least two: 44 ... Kd7 (definitely) and 44 ... g6 (probably).

I’ve yet to be convinced that 44 ... h5 45 Kd4 Kd7 46 Kd5 c6+ 47 Kc5 Ke6 48 Kxc6 Kxe5 is only a draw (e.g. 49 Kd7 Kf4 50 Ke6 g5). Now if 44 ... h5 is also a win then the presence of three solutions would mean that I would stand by my earlier assertion that today’s position, whilst fairly interesting, doesn’t really constitute a puzzle at all.

By the strictest guidelines of a problemist purist, I would agree with you. Certainly, if I were administering a competitive chess quiz, I would want to use problems with only one solution. But if CG followed those guidelines strictly we would miss a lot of good positions. I'm quite happy having the occasional puzzle that is a good learning exercise and has more than one good answer.

<However, 49 ... Kf5 is an improvement and I can't see how white saves it after that.>

It seems I really was suffering a mental block on this and needed to see some further analysis to understand just how white could save the game.

<johnlspouge ... I saw no point in being the one to reiterate my previous statement.>

The key move is not difficult and, as others have pointed out, is not the only winning move in the given position. The difficulty in the puzzle lies in some of the finesses involved further along in the ending. For example, Black must see that 51...g5+? would be a blunder allowing White to save a draw: 52.Kf5 g4 (52...Kg3 53.Kxg5 h4 54.Kh5=) 53.Kf4 g3 54.Kf5 stalemate.

I understand - normally we're used to seeing a king defending in front of an advancing pawn, so our intuition tells us that the white king is too far out of play to defend.

Think nothing of it.

44 ... h5

forces white to choose which pawns to attack with his king as his pawns are no longer effective. If he chooses the c pawn then black will take the e pawn and make it to the other two first. If he chooses the g/h pawns then the c pawn will promote...

45 Kd4 Kd7
46 Kd5 c6+
47 Kc5 Ke6
48 Kxc6 Kxe5>

You'd think so wouldn't you. I certainly did, but let me save you the bother of picking through three pages of kibitzing to provide you with the refutation, courtesy of <CHESSTTCAMPS>.

Analysis by Stockfish 9 64 POPCNT: