After learning simple piece development and basic checkmating techniques, it's tactics, tactics, tactics. The positional concepts you espouse like rooks on open files are only useful to lower level players in service of opening up opportunities for tactics. Rather than learning pawn structure directly, he recommends actively sacrificing pawns to open up tactical opportunities. He doesn't recommend ignoring positional concepts of course, but he expects his students to learn them slowly, almost by accident, during focused practice on tactics.
Modern chess engines (pre-NN) even work the same way. Traditional chess engines with limited memory and limited depth search had rich, complex evaluation functions to try to compare the positional strength of a position. Modern chess engines have evolved towards simpler, faster evaluations, instead relying on vastly deeper tactical searches with gigabyte-size transposition tables. It's all in service of tactics.
> After learning simple piece development and basic checkmating techniques, it's tactics, tactics, tactics.
Fair enough, tactics is fun, pretty, and rewarding. I like it, too, and do not advocate avoiding it. But if any of their games lasts till the endgame, it is usually the principles not the tactics that contributes to the victory.
> Modern chess engines have evolved towards simpler, faster evaluations
Alas, modern humans do not undergo Moore's law so your analogy to computer chess can as well be turned upside down: we ought to improve our evaluation function if we want to play better with the limited resources of our brain.
For example here's Jeremy Silman talking about it:
https://web.archive.org/web/20120205054644/http://www.jeremy...
After learning simple piece development and basic checkmating techniques, it's tactics, tactics, tactics. The positional concepts you espouse like rooks on open files are only useful to lower level players in service of opening up opportunities for tactics. Rather than learning pawn structure directly, he recommends actively sacrificing pawns to open up tactical opportunities. He doesn't recommend ignoring positional concepts of course, but he expects his students to learn them slowly, almost by accident, during focused practice on tactics.
Modern chess engines (pre-NN) even work the same way. Traditional chess engines with limited memory and limited depth search had rich, complex evaluation functions to try to compare the positional strength of a position. Modern chess engines have evolved towards simpler, faster evaluations, instead relying on vastly deeper tactical searches with gigabyte-size transposition tables. It's all in service of tactics.