Another potential drawback of your method comes into my mind:
Some squares occupied by one's adversary's pieces should enter the contest for ideal squares:
Namely, if they can easily be dispelled from there (without immediate tactical gain).
But that's an interesting approach!
@NochEinSpieler said in #21:
> Another potential drawback of your method comes into my mind:
>
> Some squares occupied by one's adversary's pieces should enter the contest for ideal squares:
> Namely, if they can easily be dispelled from there (without immediate tactical gain).
>
> But that's an interesting approach!
I would even say that *any* square occupied by another piece should still be a candidate for the optimal square. For example if you take the second example with the bishop from Beliavsky-Romanishin 1978, the ideal square would still be g2 even if White currently had a knight there that had to move away first.
But solving this using the author's method seems complicated because you now also need to consider where to put the knight instead. It's a slippery slope where you move from the original question "What is the optimal square for piece X?" to the much broader "What is the optimal configuration of all my pieces?".
Overall I think the proposed method is very clever and it's cool to see it correctly predict good squares which you usually don't see an engine do. However, due to the many limitations I doubt that in practice this will give you any interesting or surprising results that go beyond what an intermediate player would come up with.
Reading your article, i am now thinking that the next revolution for chess is the description of strategy by computers such that human can understand it. For now, computers can give precise (not perfect) lines that surpass any human by far. But even with the new AI developments, a good chess book is still better at explaining chess than a computer.
While reading I understood what you mean. Very nice blog post about a very important topic, I like the way you presented it. It gives you a small introduction about the topic, if you like it you can find many well written posts about it. But it's working as a good introduction because this topic is very complicated. I like the post as all other posts from you !!! Keep going!
@jk_182
Really cool idea - how can we try this software?
hey god work
ment good work
interesting post. Is your program available somewhere to download and try?
from my point of view, this is an interesting way to build chess understanding, not so much by using/exploiting stockfish, but by inspecting the strengths and shortcomings of your/any approach.
To me, one problem, that is going to stay, is the vague chess lingo, we people use, in this case:
the concept of "ideal squares"
Of course, sometimes squares are better suited than others to reach a specific goal, but even a master like Karpov, who was very good at identifying ideal squares, was very aware of the fact, that the concept relies on something that does not change (like the character of a position in positional terms), mostly dictated by the pawn structure, and maybe the king position. But many of us understand, that those things can change, just it needs more time than to execute a combo. And also some piece placement, that i would deem ideal, could well be centered around tactical motivs rather than pure positional aspects.
This goes to show, that the concept itself has to be rather vague and thus is hard to measure. And of course some squares are ideal in a sense, that comes coupled with other pieces working powers, so just isolating a single piece cannot possibly inspire the strength of a "formation". But this and many more aspects of the royal game can be discovered by "playing around" as you do with software and their limitations.
And thanks for your distinct explanations sharing your discoveries in an entertaining way.
I could see king position evaluations working well in certain types of endgames where it functions more like an attacker than a target to be protected. I imagine it could give weird results during middlegames though, not just due to the need to move other pieces first but also since the engine might come up with "white king to hole on h6"-plans in positions where such a slow king march would easily allow black time to get mating material over there in time.
Did you run this on endgames (which also often revolve around getting the pieces to the right square with the right side to move) or only for positional middlegames?