 |
| Black has just recaptured on d5 |
*****
*****
*****
*****
*****
*****
I played 52.b4, and Black replied 52...g5, more-or-less guaranteeing a draw.
Stockfish17.1 and Dragon1 want White to play 52.Kf4, which prevents ...g5 and so wins, according to Dragon1 (it gives White's advantage as almost the equivalent of a queen).
Stockfish17.1 is a lot less certain about White's advantage, but nevertheless reckons White has the upper hand.
What neither engine appreciates is that after 52.Kf4 Black has a simple draw with 52...Bxb3!, the well-known point being White's bishop is the wrong colour for queening the h pawn.
After the further moves 53.Bxb3 Kg7 54.Kg5 Kh7 55.Kf6 Kh8 the engines stick to their assessments, ie "winning" - Dragon1; "upper hand" - Stockfish17.1, but can find nothing better than marking time with 56.Be6.
Another way of spotting the fallibility of engines is to dig out old games of yours in which you made computer-checked notes using the technology then available.
Quite often I find even a cursory look at such a game of mine with modern engines will reveal what, with the benefit of hindsight, seem rather glaring errors by older machines.
Occasionally I even get the satisfaction of seeing my original move, condemned by a version of Fritz or Junior, receiving praise from Stockfish17.1 and/or Dragon1.
LESSON: popes may be infallible, but chess engines are not.
Game 5 of the Karpov - Korchnoi match of 1978 reached a similar position and Karpov played on. That was a mirror of your position but with a probably important difference that Korchnoi's pawns were one square further forward on the equivalent of h4 and g5 whilst Karpov's pawn was on h3. Thus the Bishop could defend the h pawn.
ReplyDeleteThere were presumably positions where Korchnoi could lose if he had played the wrong move.
RdC
That's an interesting comparison, especially as Dragon1 reckons White is winning, even in the final, hopelessly drawn position
Delete