Opposite Conundrum

IN my round-nine game at Alicante I had white against a Belgian Fide master.

Here I want to concentrate on the ending of opposite-coloured bishops.

It was drawn after 38 moves, but for much of it the analysis engines Stockfish14.1 and Komodo12.1.1 reckon Black is winning.

More than that, a lot of the time they reckon Black is winning by a large margin.

A critical point comes at move 50.

Position after White played 50.Kd1-e2

*****

*****

*****

*****

*****

Black is a pawn up, which often is not enough for a win in an ending with opposite-coloured bishops and no other pieces (apart from kings).

But here White has four pawns fixed on the same squares as Black's light-squared bishop, with the ones at f3 and d3 being targets for the bishop.

True, Black has four pawns fixed on the same squares as White's dark-coloured bishop, but it is difficult for the bishop to get at them.

The engines reckon Black is winning. Stockfish14.1 gives Black's advantage as worth 4.26 pawns, while Komodo12.1.1 gives a more-conservative 2.1 pawns (in both cases the advantage increases, slightly, with the more time the engines are given).

However engines are not very good at assessing endings with opposite-coloured bishops.

It is perhaps significant that in the diagram they reckon Black's advantage is exactly the same after any reasonable move - often a sign an engine's evaluation is seriously wrong.

To test this I am going to play the position out, taking white while Stockfish14.1 plays black.

50...Kc6 51.Kd2 b5

Clearly Black has to play this at some point. The key to a correct defence, assuming there is one, is to have the white bishop poised to counterattack the black central pawns.

52.cxb5+

Not much choice as Black can force this capture anyway by playing ...Be6.

52...Kxb5!? 53.Ke2

The king must not stray too far from the f3 pawn.

53...Be6 54.a3 Bd7!? 55.Kd2 Ka4 56.Ke2 Kxa3

Black has won a second pawn and increased his advantage to 4.49 pawns, according to Stockfish14.1, but now White can threaten counterplay against d6

57.Ba5 Kb3!

This is an only-move, according to Stockfish14.1, which reckons everything else allows White to draw, eg 57...Kb2 58.Kd2 gives White a superior version of the game.

58.Kd2!

The black king must be kept out of the centre, so the natural-looking 58.Bc7 loses to 58...Kxc3 59.Bxd6 Kd4 and 60...Bb5 etc.

58...c4 59.dxc4 Kxc4 60.Bc7 Kc5 61.Kd3 Bb5+ 62.Kd2 Kc6 63.Ba5

63.Bb8? a5.

63...d5 64.Bc3 Kd6 65.Bb4+ Ke6 66.Ba5 Bf1 67.Bc7 Bg2 68.exd5+ Kxd5 69.Ke2 e4 70.fxe4+ Kxe4 71.Bd8 f3+ 72.Kf2 Kd4 1-0

I resigned here as clearly the a pawn is going to cost me my bishop.

So unless a major improvement can be found for White, it seems the engines' assessment of this ending is much better than I suspected.

Arguably it is more instructive to understand how an FM failed to win  ... for which see https://beauchess.blogspot.com/2021/12/alicante-round-nine.html.