The set up for this thought experiment is simple. Someone trustworthy makes you a map to get from point A to point B in a very dense forest. There is pretty much a 100% chance you will follow the route the first time - because you are not stupid - even though it is technically possible to do otherwise.

If we reset everything and allow you to take a different route from point A to point B, you do the same thing, take the same route, and end up at B, although maybe it was easier the second time. That is the rational, safe thing to do because you cannot see if there are alternate, more efficient routes through such a dense forest.

We keep running this simulation, and eventually some things should begin to happen. For one, you will continue to make mistakes that will eventually lead to developing alternate paths through the forest; maybe you trip and fall through some brush and discover a stream running through the forest. You might also just discover an alternate way around some boulders or something.

If we run this simulation enough times we will probably get a good sense of how to navigate the forest merely by trying to get from point A to point B many times.

But maybe we could have gotten there quicker if we had just explored the forest more, right?

Well, its the densest fucking forest, and you get slapped with branches in the face for trying to deviate too far from the initial route. And it's huge. So, intentional deviations are possible, but they are often punished or a waste of time.

My question is: is simulating the navigation of the forest from point A to point B over and over again the most efficient way to discover useful knowledge of the forest itself? I would say yes, even if this knowledge is mostly only useful for getting from point A to point B.

One might argue, being some sort of bizarre treant-human hybrid, that one must know every inch of this forest as well as them to be effective, after all - they’ve been rummaging here in dead logs for fat, juicy caterpillars for twenty years.

Well, one could do that - or one could continue to learn how to navigate other similar, or dissimilar, forests, eventually developing a larger map with many different nodes that connect - a sort of nexus of knowledge that can be applied as needed in which the efficiency of traveling between those nodes becomes increasingly important.

That seems to me to be the best path to learning many, if not most, martial arts things: start with a simple jiu-jitsu sweep, for instance, go from certain specific preconditions, and then execute it such that certain (obvious) end conditions are met. Over and over. That the sweep might not have been perfectly executed matters only as iterations are performed - and even then it isn't the most important thing; the sweep itself is just a tool, not an end in itself, and viewing it that way makes it fit better into the larger context of advancing positions, or doing moves, in sequence - all of it fitting into something similar to the nexus I mentioned.

This could also apply to combining individual tactics and strategies such that they interlock or become contiguous, with individual moves making up the necessary sub-strata. Something like that could probably be applied outside of jiu-jitsu, though.

Also, I hope I'm not committing a sin by putting the metaphor first or something; that just felt like the most straightforward way of expressing myself.