The trick with the new engines is to see where they suggest sensible, human looking, moves that aren't already common theory.

Can find some quite interesting ideas that way.

That they mostly reproduce classical theory is a useful check that they aren't just producing gibberish

'Esoteric'? Hardly. 'Abstract'? Based on statistical findings, indeed. Sometimes, 'abstract' findings become the building blocks of an opening theory. Surely at the very least, what has been said above can bring about a special emphasis upon one particular line out of a considerable vast repertoire; orthodox variation. Once you increase the data it is entirely possible that a theme arises out of a mere collection of interesting 'coincidences'. 
Or perhaps it holds no merit; there's always that possibility and I welcome that discussion. And this what this is, a mere discussion, no concrete line suggestions here. As we both agreed, opening evaluations are perilous. Nevertheless, I do completely understand if this sub-forum concerns itself with specific line suggestion, you have my approval of locking/delete this thread as you please (I saw no delete option). 

tony37 wrote on 06/02/21 at 13:46:06:


You are quite right,that would be interesting yes....but we are moving into a different area; what I do to satisfy that question is to create polyglot opening books from kings indian-exclusive databases, then run stockfish13 engine tournaments whereby one is locked in with the opening book and one is free to do as it pleases. Any conclusions are derived only after 100+ games.   Having said that, TCEC (Top Chess Engine Championship)also has in its opening books for engines both kings indian defence and attack..matter of fact, the tournament that runs currently uses them quite regularly. So a collection of those games  would partly answer what you are addressing here.

Bibs wrote on 06/02/21 at 13:41:04:


Thank you for the welcome. And yes, this is exactly what i find interesting; when independent engine computation arrives on a previously held human theory that was derived through the stages of history, trial and error. Now it may happen all the time or it may not , but when it does to my favourite opening...i just liked to point that out   

This also explains why the line is so short; this is what happens when you let the engine lead the way after the 13th ply.

ibb.co/1qKJTMq is a picture of the situation in excel ->The first six top options for move 7.5 and onward are displayed. Numbers colourfully attached to engine moves are  ordered in the following way; evaluation(99.1--100% unless max duration ),depth and lastly at the bottom row is duration of computation (minutes,max=21).   

Conventional lines of the king's indian are identified with specific opponent moves. They are inseparable..once you start entering different values, you are deviating away from theory. Which does not serve the purpose here; to find coincidences between engine computation and historic human theory.. if you have any examples do share.

Welcome!

To be fair though, strong as engines are, it's not obvious that running engines at whatever depth will generate all that much usable new insight in the first few moves.   

But yes, kinda interesting to see what the NNUEs 'decide' to play after a few million walks around any particular park.