Here's the position just before 98...Rg6:

click for larger view

White's last move was 98. Kc5. This unpins the bishop and intends Kd4. When the white king arrives on d4, he protects e5 and the bishop is free to move.

Then white has a sneaky threat. If allowed he plays Qf8, Be7 and Bf6 threatening mate on g7. It doesn't matter if black plays gxf6, as white would just recapture with a pawn and the mate threat stays in place.

That possibly explains why black plays 99...Rg6. He needs to protect g7 before white locks him out with Bf6.

Fritzie doesn't find a magical escape for black. He reckons that the best choice is 98...Bf5 (+6.10), intending Bg6, Bh5 and Rg6 (anyway). 98...Rg6 is not far behind with (+6.57) followed by 98...Kg6 (+6.87), 98...Bg6 (+7.07) and 98...Bd3 (+7.5). And then we are into the dross.

So can we answer our questions? I think so. It seems that however black plays he loses pawns and the fortress crumbles. 98...Rg6 may not be the best move in the position, but it is not far from it, and it takes 21st century software to spot anything better.

When I first played through the game (quickly) it looked as if white won by having more stamina and out-sitting Samisch until he made a mistake through tiredness. But a deeper look shows that both players had seen quite a bit deeper into the position.

The fortress works like this: If white takes the "a" pawn, then checks black on the back rank, black just moves his king to h7 and leaves it there. While white's queen remains on the 8th rank, all the second rank squares are uncovered for black's rook to move back and forth freely. If, however, white moves his queen off the 8th rank, then black re-plants his rook on a2 and the moves his king back to g8, and shuttles to h8 and basically tells white --- whaddya gonna do?

If white pushes the g pawn, black always plays fxg4. If white then recaptures and eventually pushes g5, black always plays hxg5. If white tries to prevent that by moving his queen to the newly-opened h-file while black's king is on h1, black then moves back to g1 whenever white's queen occupies the h-file, preventing any pin.

(The key is for black to NEVER move his bishop, because doing so would eventually allow e6!, opening the black-square a1-h8 diagonal for white's bishop, possibly allowing tactical shots.)

So, after all this, it seems there is only one conceivable plan for white to break through. Imagine from the position on move 41 all these pawns off the board : a6 f5 g3 h3 h6 -- all gone. The only plan for white is to get the queen on the g-file, the black king on g1, and the white bishop then moving from e7 to f6, finally setting up a mating net. But can't black always avoid that because once the pawns are all gone, his rook is a free agent and he should be able to get it to c3-h3-h6, and then shuttle back and forth from h6 to g6, preventing any mating attempts on g7?

Perhaps there is a fissure in this fortress somewhere, but I don't see it.

Why not simply playing 98. Bg6 (or Bf5) instead of 98...Rg6? The reason is - as already pointed out by <Once> - the weak point g7:

98...Bg6 99. Qf8 Bh5 100. Kd4 Bg6 101. Be7 Bh5 102. Bf6 and Black is lost

click for larger view

Or 42...Ra8 - a horribly passive square for a rook. Then white invades on the queenside with moves like Qa4-Qc6 or Qa5-b6-b7 (depending on how black tries to keep him out by moving the bishop). Eventually the black rook will have to move away from the a file and the pawn will fall.

A lesson to all the club players who would resign if they were down R for Q going into the endgame.

I must confess that I hadn't considered the setup a6/Rc6/Bd5. That's probably because I instinctively don't like it. You see, I have grown to mistrust what I call PPP - pieces protecting pieces. If you have a chain of pieces with one protecting the other, your position becomes inflexible as nothing can move very far. And you can be vulnerable to deflection tactics.

Having said that, Fritzie quite likes your idea. He rates 42...Rc4 as +3.13 and Rc6 as +3.23. Maybe the point is that white has few pieces left to disrupt black's chain of pieces.

Good call!

Meanwhile Black could have fought on for a bit in the final position:

click for larger view

with 101. Rxg5. Crafty End Game Trainer link (a win starting Qe8 may be there but is not easy to find): http://www.chessvideos.tv/endgame-t...

Then white has a sneaky threat. If allowed he plays Qf8, Be7 and Bf6 threatening mate on g7. It doesn't matter if black plays gxf6, as white would just recapture with a pawn and the mate threat stays in place.>