Patterns and Calculation

There has been quite a bit of discussion concerning pattern recognition the past week on Chess.com. An article by a philosophy student,* "Pattern Recognition: Fact or Fiction?", provoked several dozen comments, many challenging the author's analysis. There is also a forum thread that spins off this article, "'Pattern Recognition' DEBUNKED", and another thread on the topic in a closed group for over the board players. The private group's thread started in August and inquires into the practicality of creating a pattern bank. Would it need 10,000 positions? More? I have contributed to all of these threads.

These discussions reveal an absence of a clear and accepted definition of patterns in chess. Are patterns a static arrangement of pieces that crop up with some regularity? Are patterns dynamic relationships, such as all pins constituting either single pattern or perhaps a specific category of patterns? What about typical pawn structures, such as the Caro-Kann structure that also commonly crops up in the Scandinavian Defense (see Panayotis Frendzas' review of Vassilios Kotronias, The Safest Scandinavian)?

These questions linger in the back of my mind, becoming active while reading a chess book, solving tactics problems, or playing. I am currently reading with an aim to reviewing Paul Powell, The Fighting Dragon: How to Defeat the Yugoslav Attack (2016). Powell makes pattern recognition central to his approach to opening study. My last youth lesson before the holidays focused on a simple checkmate combination that occurred in a blitz game and is part of my Knight Award tactics set (see "Pattern Training"). Next week at a chess camp, I am teaching a class on the Qh6+ sacrifice that ended this year's World Chess Championship match. Sam Copeland created a video on the topic for Chess.com. My work begins with his challenge to find more examples of this pattern.

This morning I solved two tactics problems on Chess.com's tactics trainer. The first one had a 2002 rating but took me a mere sixteen seconds. I had seen the same problem a few days ago and spent several minutes calculating before solving it successfully. When I saw it this morning, I recognized it after about ten seconds. Instantly, I knew that I had to attack the queen with my knight. A few seconds were needed to either remember or quickly recalculate the correct square among the two possibilities.

The second problem gave me more difficulty.

White to move

Naturally, I quickly looked at 1.Nxd1, rejecting it in the light of the fork of knight and pawn by 1...Rd4. It was clear that I needed to push my pawns, but experienced a good deal of confusion about how that was possible. Not only did it seem that the rook could stifle the ambitions of either pawn, but also I quickly saw that 2...Rxa8 or 2...Rxd8 would be checkmate. I spent some time calculating lines that begin with 1.Kg8 with the idea to support the d-pawn. These fail.

After about six minutes, I realized the rook was overworked and knew the first move.

Slowly a learned pattern emerged in my memory. Two connected passed pawns on the sixth rank are too much for a rook. But, my pawns are separated. Nonetheless, a solution dawned on me! 1.d7, pushing the pawn that the rook is not behind (as one would do if the pawns were connected). 1...Rd4 2.a7.

The rook cannot stop both pawns! But Black has another resource.

2...Bf3

White to move

Both promotion squares are guarded. Time to calculate further. Earlier, during my confusion, I had looked at Ng4+, seeing that it led nowhere. But, now, this move decoys the bishop from protection of the promotion square.

The decoy theme is certainly a dynamic pattern.

Of course, the king can move out of check, so the bishop will not be distracted so easily. In my calculation, I began to comprehend why the problem composer put a pawn on h4 (I'm assuming the problem is composed).

During my calculations, another pattern revealed itself: interference. 3.Ng4+ is the correct move! 3...Kh5 (3...Kg6 allows 4.Ne5+ forking king and bishop) 4.Nf6+ Kxh4 5.Nd5!

Black to move

If the bishop captures the knight, the rook no longer guards d8. If the rook captures the knight, the bishop no longer guards a8.

This problem could have been solved by pure calculation. That it took me more than ten minutes to solve, suggests that calculation was my best resource. Even so, along the way, patterns that were not instantly clear to me guided me and aided the calculation.

Because the problem took me so long, I gained only one point on my tactics rating. The average solving time is 2:34, but 2/3 of those who attempt the problem fail.


*He identifies himself as a teacher who was trained in philosophy. Readers of Plato understand that Socrates always thought of himself as a student, as a lover of wisdom who pursues knowledge and truth.