Clever chunking

by Collin Lysford

An example surfaced from Nintil’s Scaling Tacit Knowledge, about memorizing chess games:

I would have considered it a waste of time before giving it a chance. The trick is that to memorize a game, you sort of have to understand it. It’s possible to just memorize moves like you’d memorize a list of random words, but it will be 10x harder than just understanding what’s going on.

If you understand what’s going on, you end up memorizing the game in a series of chunks, instead of a series of moves. For example one miniature that I’ve memorized is this game between Peter De Bortoli and Botond Smaraglay. I can recite move by move, but the way I remember it is roughly Smith Morra gambit, knight development, bishop development, scare off bishop, threaten queen trap, knight blunder, queen trap”. Memorizing a couple king’s gambit games has definitely improved my king’s gambit play by giving me more ideas.

Why does chunking makes things easier to memorize? The same reason words are easier to remember than arbitrary strings of letters - because we use them. Knight development” is a pattern of chess moves that are far more likely than random chance to happen in sequence, because they lead to an unusually good outcome (relative to the space of all possible moves) when executed. It’s memorable because it’s common, it’s common because it’s useful, and it’s useful because people often like what happens when they do it.

I don’t actually know what constitutes knight development”, because I don’t really follow chess. The extent of my knowledge is that your knights are more free to move and threaten enemy pieces, but I couldn’t tell you what exactly constitutes knight development”. But I don’t need to know that to understand what’s happening when I see a chunk like knight development”. Almost every domain of any complexity has this same phenomena of fluid ontology, where chunks that aren’t based on the root-level objects (one piece, one turn) get parceled out by language owing to their uncommon usefulness.

You need chess experience to come up with knight development” as a useful way to describe a series of chess moves, but you don’t need chess experience to recognize knight development” as a chunk derived from experience. Domain-specific language that’s been matured through experience just looks a certain way, and that way can be recognized without deep knowledge of that domain. (An analogy: you need specific botanical knowledge to look at a single tree and estimate how old it is, but you don’t need that same knowledge to tell that it at-least-so-old.) You can use this to jump-start your understanding in a new field by instantly identifying sources that are far more likely to be credible owing to the relative maturity” of their chunking, as expressed through langauge. I think people are sometimes hesitant to admit this, because it sounds like a sort of petty elitism or classism - I can tell from the way that you talk that you’re more likely to be correct, even if I don’t know anything about what you’re talking about.” But, well, sometimes you can tell from the way people talk that they’ve spent a lot of time interacting with the domain - and if that doesn’t make them more likely to be correct, what the hell does?