Think Like a Grandmaster (Kotov's connections to Cognitive Science)

My return to regular chess-playing has been accompanied by an interest in revisiting some of the chess books I read (ok, maybe it's more fair to say "looked at") when I was younger. Something I've been looking for in these classics besides positional insights and tactical advice are hints regarding the cognitive mechanisms that expert players bring to bear during play. Chess is a hobby for me, but cognitive science is my career - I'm a vision scientist by training, with research interests focused on visual recognition (faces and textures in particular). I'm also broadly interested in what are sometimes called visual routines, which refers to processes we use to understand spatial relationships visually: How do you work out a path through a maze, for example, or how do you work out which parts of a complicated diagram are connected? This latter topic has a lot of interesting overlap with chess and in the spirit of blending work with play as much as possible, I've been looking around for what connections others have already made between playing chess and the study of the mind. I've got a pretty good stack of books at the moment (I've got a TON to say about the intriguing work of Adriaan de Groot documented in his book Thought and Choice in Chess in particular), but one of the first I decided to tackle was Alexander Kotov's Think Like a Grandmaster. With a title like that, there's just gotta be some good cognition in there somewhere, right?

Lots of writers and chess players have tried to describe the psychological aspects of playing chess and I've been working on familiarizing myself with some of their ideas. One thing I'll say right away is that few of them have much to say with what I'd describe as the cognitive science of chess and are instead more oriented towards the psychology of chess. That distinction may not seem like a big one to you (or some of my colleagues in the academy), but to me it matters a great deal. The former discipline - cognitive science - is specifically focused on thinking about the mind in computational terms, with sub-domains concerned with perception, memory, attention, and decision-making and points of intersection with computer science, graphics, and philosophy. The latter is a broader field, which includes topics like personality, social interaction, and other aspects of human behavior that are perhaps not so easily discussed in terms of algorithms and other explicit models of performance.

image

Why do I say a lot of chess writers are more focused on psychology? A lot of them want to tell you a lot about the various habits or nervous tics of different players, for example, or encourage you to study your opponent in an almost Freudian way (Does the player fear open positions? Are they temperamentally equipped for a long positional fight?). I think a lot of these discussions are sort of seductive because they appeal to the characterization of chess as a battle of wits: Besides the objective battle happening on the board, the idea of understanding your opponent psychologically adds a less clear-cut dimension of social engineering to the game. Can you psyche your opponent out somehow? If you've researched their quirks, can you leverage that knowledge to push them into situations that you know will make them nervous and dominate them that way? I still encounter a lot of advisees in my Psychology department who love the idea of being a criminal profiler and I think some of the same things that make that career sound glamorous are at play when chess writers and readers pore over anecdotes about walking around during play vs. staying seated, or making moves with subtle gestures rather than big sweeping movements. For me, however, this isn't all that satisfying. What I'm looking for are ideas about the cognitive processes that support good chess playing - how are perception, memory, or attention trained and applied to the game? What do good players do that novices don't?

Luckily, Kotov has some neat stuff to say about that, although some of it is a little buried in the text. Besides the exercises and the general advice about different kinds of positions, I found that every so often he'd say something in a sort of off-hand way that struck me as a potentially profound statement about the intersection of cognitive science and the game of chess. I'm not going to go into these topics in depth right now - instead, I want to highlight some of the ideas I found the most interesting to bookmark them for later posts. Some of these might be hypotheses that various studies in the cognitive science literature have examined, while some of them might be totally unexplored (or totally wrong!). In no particular order:

1. Why do chess players (even good ones) sometimes miss obvious replies by their opponent? - It happens all the time: You're working on your next move and you've come up with a good idea. You think about the different things your opponent might do and they all look like they lead to a good outcome for you. You wait, you think again, you nod to yourself and make the move...and your opponent promptly does something you didn't anticipate at all! Suddenly you're toast and you have no idea how you didn't see that they could just push the pawn, or recapture with the knight instead of the bishop, or threaten your queen while you attacked theirs, and on and on...

There are a lot of things to explore here I think, but the first thing it made me think of was confirmation bias, which looms large in many discussions about reasoning and decision-making in the cognitive science literature. Briefly, confirmation bias refers to the tendency to search for evidence and/or process information in a manner that supports what you believe. To put it another way, we tend to be quite bad at closely examining evidence that might prove us wrong about something. In the case of a chess game, it's obvious what the prior belief is: This move is great! Confirmation bias in this setting means that we don't go out of our way to look for responses that make the move a blunder, but instead focus on the various ways it could work out well for us.

A simple diagram of how confirmation bias leads us to ignore data that could contradict our beliefs. (from SimplyPsychology.org)

A diagram showing off how confirmation bias can lead us to ignore data that could contradict our beliefs. From SimplyPsychology.org. (link here)

There's a phrase I've picked up from watching IM Levy Rozman (GothamChess) that seems relevant here: "Hope Chess." This refers to playing moves and hoping that your opponent will do the thing you want rather than make the move that will leave you stuck without a path forward. Hope chess and confirmation bias are tightly coupled, and Levy's advice about how to avoid hoping in favor of playing is pretty sound - be systematic about how you consider moves and look to answer key questions like "Can they check me?" and "Can they capture something?" as specifically as possible. It would be neat to hear more about how strong players adopt strategies to mitigate confirmation bias (if they do).

There's another aspect of cognition potentially at play here that also could be very interesting to explore, though: The limits on selective attention. Briefly, the sensory world is rich and complicated, making it impossible to measure and understand every stimulus impinging on our nervous system all at once. What our minds and brains do instead is filter, or select, some subset of the environment to work with. The good news is that we can prioritize some stimuli for further processing, but the bad news is that we can completely miss the things that we filtered out. You may have experienced an extreme version of this if you've encountered one of Dan Simon's classic demos (you can watch one here and I won't spoil it for you!), but in ordinary circumstances it also happens all the time in subtler ways. How does attentional selection work over the chessboard? Do GMs have a wider attentional spotlight and if so, how did they get it? On the other hand, do they use other mechanisms to maintain good board vision to compensate for the limitations on their attention? All neat questions that I'm not sure we know much about.

2. Intuition vs. Analysis - Kotov spends a lot of time discussing how to properly calculate in chess positions, but also emphasizes that a strong player also uses intuition or judgment in many cases too. That is, rather than explicitly describing a series of moves that leads to a desired outcome, strong players often rely on pattern recognition - the ability to look at an arrangement of pieces and understand things like who is better off, which squares are strong or weak, and what kinds of things are likely to happen in the long run (e.g. "The kingside pawns will push through eventually.") I'm especially interested in this distinction between intuition and analysis because it intersects with my work in visual perception. Intuition sounds a lot like learning to recognize and label different visual stimuli, which is something I think about in my research in a bunch of different ways. A key question for me with regard to chess is what exactly is being recognized? Is it just a configuration of pieces? Is it instead the spatial transformations that a position supports?

This leads us into another idea Kotov brings up a few times, which is the difference between the static features of a position and the dynamic features. Static features are things like the material imbalances or the current positions of the pieces. Dynamic features are a little harder to define, but they refer to the ways in which pieces will be able to move either singly or in a coordinated way to transform a player's control of space, or to mount an attack on some part of the board. Kotov borrows language from physics about potential energy and kinetic energy to talk about static and dynamic features, and I'm really interested in whether and how these may be different kinds of visuospatial processing. If you watch chess streamers, a lot of them use arrows to think out loud, which suggests to me that there is perhaps an independent spatial form of processing that's maybe different than the static evaluation of what's in front of them.

3. Write moves down, don't analyse together, etc.
Besides these bigger concepts, Kotov also gives us a few nuggets of wisdom that fascinate me because they are so specific. For example, he endorses the idea of writing a potential move down on your scoresheet before making it as a way of kicking yourself out of the confirmation bias rut. He also (to my surprise) suggests that group analysis of a position often goes worse than individual analysis. Take that Wisdom-of-the-Crowd! If you aren't familiar with that term, it refers to a phenomenon in which the combined judgments of a large number of naive observers can be more accurate than the judgment of a single expert. Interesting to hear how this may play out differently when you have a bunch of experts all considering the same position. There is also an interesting digression about the value of walking during a game instead of just staring at the board. A bunch of these seem pretty easy to test, but I have no idea if anyone has! They aren't all necessarily the most profound ideas about chess and cognitive science, but I'm curious to see what we know from the empirical literature.

I'll probably find more in his book after I read it again, but already it seems like there are a lot of neat directions for hunting through research articles and monographs. I'll be posting more soon about some of my other thoughts and explorations in this domain, so keep an eye on this space if you're interested in what chess has to tell us about cognition and vice-versa.