In the diagram position from a 2018 Annex Club tournament game Black played 1 … Bd1 completely missing the point that 1 … Be4! Is crushing. The reason being is the Bishop and King totally dominate the misplaced Knight’s squares. The Black King can rush at the beleaguered White g3 pawn.

Often Bishops can dominate Knights in an endgame in this manner. White in this game wasted no time in redirecting her Knight via 2. Ng2 and 3. Nf4. She went on later to Queen her eventual White pawn on h7 by cleverly moving the Knight to c6 to block out the Black Bishop from the queening square on a8. The Knight got its revenge.

These are moves that even Masters can miss. One of the things that gets our juices going at CIC is the difference in human thinking from computer programs. I once beat IM Zvonko Vranesic in a simul at Hart House. Zvonko gave up his Chess career to become a professor of Computer Science at University of Toronto. During an interview, he said that computers have not captured the human decision making process. Often winning moves like this are beyond the move horizon for even the best Chess software, or are too subtle for reducing to an easy algorithmic approach. So, programs with a brute-force ply approach can struggle.

Recently at the London Chess conference, we learned directly from the AlphaZero team how this system has a different, more human-like approach. AlphaZero uses its neural network over thousands of games to learn rule-sets. It then looks at fewer positions but in far more depth. AlphaZero would have found this move easily i.e. “cut down opponent’s Knight mobility.”

Chess is not only fun, it is a great tool for exploring human thinking, and what may be some of the future avenues in artificial intelligence. This is the sort of key research that I personally hope Chess Institute could one day support.

And you thought this was a boring ol’ game.

John