When I pick the combination of the day for ChessPuzzle.net from tactics that my algorithm generates from game files, I always look out for surprising solutions. „Think like a Mashine!“ has given me a much better idea what this entails. It’s not an easy book, definitely not for novices. The combinations are really advanced. This is by necessity. The authors, Noam Manella and Zeev Zohar from Israel, have been looking specifically for tactics that even strong grandmasters fail to see.
Puzzles work differently than actual games. When you solve a puzzle you know that there is a tactic to be found. When you sit at the board, you don’t get a prompt (unless you cheat). Combinations get overlooked by strong players, when another move is obvious and natural. For instance a recapture. Or the capture of an unprotected piece. This much I knew. “Think like a Machine” has systematized what is normal to overlook. It is not about mistakes due to tiredness, nerves or time-trouble, nor about positions that require really deep calculation power.
Manella and Zohar have assembled positions where computers pointed out missed tactics and categorized them in seven chapters. The first is called “Against the Instinct”. It is when humans make the natural move and miss a much stronger one. Like the above mentioned recaptures. Hard to see are backward moves in order to make a threat or prepare an attack. Or quiet moves in situations when there are threats everywhere or pieces are hanging.
Chapter two is called “Total Chess”. Combinations that involve the whole board or at least different parts of the position, are hard to find. When the board is on fire you often have to weigh between attacking and defensive moves, and your direction of thinking has not only to consider White and Black but also attack and defense at the same time.
Hard to see are backward moves in order to make a threat or prepare an attack. Or quiet moves in situations when there are threats everywhere or pieces are hanging.
“Beyond the Horizon” is the topic of chapter three. We know the „horizon effect“ from engines. It used to be a problem in the past when the search was limited to a certain number of moves, and sometimes the computer would give up a pawn in order to make the loss of a piece or a checkmate disappear beyond the horizon. The authors adapt this term to human thinking. Humans tend to reject an attacking idea once they see a defense and don’t try hard enough to refute this defense. It turns out that diagonally backwards moves are especially hard to find.
Chapter four is called “Spectacular Attacks”. Sacrificial moves without capture are often overlooked. If a sacrificed piece can be captured in several ways, the calculations grow over the head of many a player. Chapter five covers “Cold-Blooded Defence”. This is actually a reference to “Practical Chess Defence”, a 2006 book by Jacob Aagaard, who is the publisher of Manella and Zohar. Defending by tactical resources is challenging. Everyone who commentates games is aware, how often such resources are pointed out by the computer. “Endgame Fantasy” is the title of chapter six about surprises in the endgame. Endgames are less a matter of „technique“ than calculation.
While all previous chapters show motives that strong human players failed to see and computers found, “Machine at the Board” is about „unhuman“ moves that were found by humans nevertheless. I assume that Manella and Zohar wanted to end on a positive, hopeful note, since they maintain that we can learn from the engines. Even though I am not convinced about that, I love their book.
Solutions: Gashimov could have won by 37…Qb1+ 38. Kh2 Nh3 (threatening Qg1+ and Qh1 mate) 39. f3 Qg1+ 40. Kxh3 Qh1+ 41. Qh2 g4+ netting him the queen. Saric missed 33. g3, and avoiding checkmate would have cost Black a rook.
Martin Bennedik was profiled here and speaks at ChessTech 2020 in the Sunday session Artificial Intelligence meets Big Data.