[from Dec-08-06]
<RookFile: *** Something that not enough people talk about is 24. Rc1. That was a mistake. <The moral of the story is: if you want a draw, keep the rooks on the board. It's by far the safest way to draw.>>

Chess is not so simple that it can be reduced to the application of reliable maxims (such as “all rook endings are drawn”). There are exceptions, such as this game: Alekhine vs Reti, 1922 In which Reti realized that exchanging rooks (with <57. … Rxg4!>) would assure him of a draw from this position (black to move):

click for larger view

For example, extremely interesting is the position in the second diagram below (which did not occur in the game but could have arisen if Cohn had played <35. fxg4!?> instead of his actual <35. e4?!>, which was routinely losing) in this position:

click for larger view

After the further hypothetical continuation <35...hxg4?!> (Better here - and more likely to have been played by Rubinstein - is <35. ... fxg4!>, but players of less-than-immortal stature would probably find the capture "towards the centre" more natural. The position after <35. ... hxg4?!> is still theoretically winning for Black, by the way, but the technical burden is much greater than after <35. ... fxg4!>.) ... <36.Kg1 f4 37.exf4 exf4 38.Kh1 g3?> (Correct is <38. ... f3! >, the only move that wins by force, as demonstated in analysis by IM Steve Giddins presented in the source cited infra.) <39.fxg3 fxg3 40.hxg3>, the resulting position would have been:

click for larger view

The position in the diagram above appears to be winning for Black due to the much stronger position of his King. Averbakh, among others, gave the line reaching this position as “ ”. Nevertheless, White surprisingly can hold, as follows: <40. ... Kxg3 41.Kg1 Kf3 42.Kf1 Ke3 43.Ke1 Kd3 44.Kd1 Kc3 45.a4! a6> (No better are either (A) <45...bxa4 46.Kc1=>; or (B) <45...Kxb4 46.axb5 Kxb5 47.Kc1=>) <46.axb5 axb5 47.Kc1 Kxb4 48.Kb2=> producing the following drawn position (with Black to move):

click for larger view

(*) - <Note by PP: - The referenced "Mestel lines" demonstrate a draw after a hypothetical <35. fxg4 hxg4 36. Kg1 f4 37. exf4 exf4 38. Kh1 g3?> [In this line, only the "study-like" <38. ... f3!> wins.>

Although <34. … e4!> is unquestionably the simplest win and therefore objectively better than Rubinstein’s <34. … g4>, the above comment exaggerates the defects of Rubinstein’s choice at move 34.

The position only would have become “study-like” if

<A> Cohn had played <35. fxg4> (instead of his actual <35. e4?> – played one move too late; although inadequate even there, at move 34, the advance of the e-pawn would have been White’s best defensive try);

<and>

<B> Rubinstein had played the second-best (but still winning) <35. … hxg4>. In that case, at move <38>, Rubinstein would have been required to find the “study-like” <38. … f3!> instead of the obvious (but inadequate) <38. … g3?>.

click for larger view

IM Steve Giddins (op. cit. [supra, in separate post from earlier today]) writes simply, “White clearly loses the e4-pawn.” (at p. 17)

This is a bit oversimplified. White could have tried <39. f4!?>, but it would have been to no avail, as these lines demonstrate:

<39.f4!?> [or 39.f3 g2 40.f4 exf4 41.e5 Kg3 42.e6 f3 43.e7 f2#] <39...exf4 40.e5 f3 41.e6> [or 41.Kf1 g2+ 42.Kf2 (42.Kg1 Kg3 43.e6 f2#) 42...Kh2 43.e6 g1Q+ 44.Kxf3 Qe1–+] <41...g2 42.e7 Kg3 43.e8Q f2#>, with this final position resulting:

click for larger view

we get the position M1:

click for larger view

White is left without a single move, complete zugzwang and time to tip the king. I saw this before I ran it through the engine.

Here..

click for larger view

23.c1? was a blunder of 1600 5min blitz levels because all you need to do is count, you don't really have to work out variations and lines or be a Grandmaster or even an expert, if you know the rules and understand chess it's just obvious. If you can count to 8 you <should see> you lose by playing Rc1 and offering the exchange, this is when playing for a draw backfires..

"+0.44 (23 ply)" means the computer searched 23 half moves deep "(23 ply)" and the optimal final position is advantage to White ("+") of 44 centipawns or a little less than half a pawn ("0.44").

Similarly, "-0.59 (28 ply)" means 28 half moves deep, advantage to Black of 59 centipawns.

<By the way, where did you see those?>.

In the "Bent Larsen's Good Move Guide" this is said that:

<The position was reached after an error of judgement by white 24.Rc1?? Rxc1 25.Kxc1. Rubinstein had foreseen that this pawn endgame was won. In this position the winning procedure can be worked out exactly. It starts with 25... Kf6!!, and the plan is to attack the isolated h-pawn and thereby force the white king to g1. After that some pawns are played forward for exchange and then the board is 'cleaned' so that the black king can go 'west' from h3, and take a pawn, perhaps on e3 or e4 or perhaps right over on the queen's wing.

Analysis by Stockfish 120719 64 POPCNT:

1. -+ (-4.52): 25...Kf6 26.Kd2 Kg5 27.Ke1 Kh4 28.Kf1 Kh3 29.Kg1 g5 30.b5 f5 31.Kh1 h5 32.Kg1 e5 33.Kh1 e4 34.fxe4 fxe4 35.Kg1 h4 36.Kh1 Kg4 37.Kg1 h3 38.f4 gxf4 39.Kf2 fxe3+ 40.Kxe3 Kf5 41.Ke2 Kf4 42.Kf2 e3+ 43.Ke2 Ke4 44.a4 b6 45.Ke1 Kd3 46.Kd1 Kc3 47.Ke2 Kb3 48.Kxe3 Kxa4 49.Kd3 Kxb5 50.Kc2 a5 51.Kb2 Kb4 52.Kb1 a4 53.Kb2 b5

1.Black to move. Last: 25.Kxc1 ~ Puzzle: p30 / Solution: p38

the plan is to attack the isolated h-pawn and thereby force the white king to g1. After that some pawns are played forward for exchange and then the board is 'cleaned' so that the black king can go 'west' from h3, and take a pawn, perhaps on e3 or e4 or perhaps right over on the queen's wing.