Chess memorization as seed planting:Relax and let the roots grow

Over a year ago, when I posted my new method for learning a bunch of tactical problems, Haunted Knight made the following excellent comment, a sentiment that has been expressed by many bloggers in response to the Circles program of study:

But something that worries me about repeating the same set of exercises over and over is that I'll just be learning that set of exercises, rather than tactics in general.

I waited all this time to reply to Dean, wanting to finish the Circles and better judge their effectiveness. I think now I have a slightly better perspective, being done with the Circles for quite a few months.

My quick and dirty response is an analogy. Memorizing a set of tactical problem or master games is like planting a bunch of seeds in a garden. Initially these seeds are isolated, unconnected. After a while, these seeds have become a garden with a beatiful complicated and interconnected root system. The difference is, for chess the roots are neurons and they can talk to each other. This is very good for us.

I said at Phaedrus' blog (when he made a similar point to Dean's):

I'm not convinced simply "memorizing" 1000 positions is all that bad. It all depends on how our brain treats those memories once they are implanted. The brain may (with no conscious effort on our part) integrate these different memories into more general categories, form cross-links among categories, striving to build an ever-more coherent picture of the chess world, even while we sleep our brain probably does this. If this speculation is right, the individual problems are like nodes in our brain that are initially implanted, but connections are formed among these nodes so ultimately it becomes a more general and useful integrated tactical skill set.

From my experience with the Circles, I think we can eliminate the worst-case scenario, in which case you memorize the exact position, and that's all you've learned so you only recognize it when it appears exactly the same as during training. This would be fairly useless. Luckily, this isn't how my memories of the solutions work. Back-rank mate, for instance, pops out at me regardless of the exact location of the King (e.g., queenside, kingside) and whether there are two or three pawns hemming him in.

In general, it seems the way our neuronal pattern recognition machinery works is to store not just the exact template, but a more general category into which it will place similar but not identical instances. For instance, once I've acquired the memory/pattern of a person's face, it then generalizes so that I can recognize him laughing, frowning, talking, at sunset, in artificial lighting, or even if his face is upside down (though in the latter instructive case it will take me longer, and I'll have to double-check just to make sure it is really him, just as when my recognition machinery kicks in for chess it is imperative to check to see that the tactic will really work given the particulars of the situation).

We don't know how this works, but we certainly don't consciously force our brain to do this. But we can probably help our brain do it by taking a single position and studying little variations on it from multiple perspectives. Like looking at an elephant from the front, you might not recognize one you see from the tail-end, so you want to walk around the elephant, see it moving around and interacting with things, to really build up a more perspective-independent ability to recognize it. This should help with basic chess tactics. E.g., study the back rank mate twenty different ways and it should stick better.

So, in contrast to Tempo's focus on the importance of conscious feedback and narratives, I give our brain more credit. I say, be more Zen. The vast majority of information processing going on in our brains is not consciously accessible, the brain does amazing and wonderful things with "isolated" individual things we've learned. It seems to strive to build models of the world, models that will generalize so we can use them in novel but similar situations. There is a tendency to want to force oneself to improve at chess via conscious exertion of will, but my hunch is that most of the learning goes on beneath consciousness, when we are sleeping, when the brain is consolidating into long-term memory the bits we have most recently learned.

Of course, we still need to work, and all the narratives and such are very helpful (in practice, I learned a bunch of tactics much faster when I started using narratives to explain each problem to myself). The two approaches are not in contradiction. Like with a seed, you can create good conditions for it to grow (lots of feedback, narratives, conscious effort), but then you have to step back and let it grow. If you fuss with it too much, try to force yourself to improve faster than is humanly possible, you will just get frustrated and may even kill the plant.