In my recent chessclub training our IM wanted us to evaluate Positions. Every position has been shown for 3 minutes, then we needed to choose from 7 Options:
Big advantage for black
Advantage for black
Slight Advantage for black
Drawish
Slight advantage for white
Advantage for White
Big advantage for white.
It were exclusively positions where there were no immediate winning tactics. It was strategical and more or less complicated.
This helped me extremely to train my feeling if i can go for initiative or being carefully. You need to visiualize where this positions leads into and therefore you can recognize what to do at best.
A trainer like this on lichess would be awsome.
So to solve a puzzle on this trainer you needed to click on the right evaluation choice and afterwards even needed to find a good move which proves that you understood the position in the longer term.
A perhaps so called strategy trainer!
Who shall create these puzzles? Chess engines are relatively weak in judging positions, so a strong human player would have to do this work.
Normally Top chess engines can at least say more about a position than human do. The puzzles needed to be solvable for chess engines. No too complicated stuff.
@sheckley666 If the positions are solvable by most players, the top chess engines can certainly solve them. Otherwise, it is too hard.
Engines are strong in tactics, but
@Elite-Tarkan wanted positions without tactics. This is the kind of positions, where engines are weak.
Of course, a top engine will find tactics in positions, where top humans believe, there aren't any tactics. But then again, the engine suggested solution would be a tactical riddle, and not a positional.
Sheckly normally a top engine can evaluate a strategical position good enough to give at least the direction to the 7 options in the request post. Just try it out - grab you some middlegames for your recent games and after you analyse them some moves ahead in its best played way, the engine gives a good advice and estimation
The idea that engines are strategically weak is outdated, and when discussing top, modern engines running on modern hardware, it's just plain false. It's certainly true that the engine does not formulate a strategy, but given good search tree culling heuristics that enable searches on optimal paths 25+ moves ahead, "strategy" simply falls out.
Fundamentally, the winning strategy in chess is to play the best move every turn (this is probably a drawing strategy when employed accurately by both sides). Engines are far, far better at this than people, which is why engines beat people even though they're "only" employing tactical analyses.
Lets say it's possible to filter out the "strategical positions" reasonably well (no idea if its difficult or not. At least not entirely trivial)
Youll get the correct evaluation (a lot of 0,00s) but have to understand the reasons for yourself (no IM to guide you through)
All i can say is that the examples i made in the club training were good to evaluate if you look close enough. It were positions where you need to recognize what is going on. Then you needed to find a move if it would be blacks AND also whites turn. Iam sure this is possible, but how to create such puzzles in a big amount is a question where iam stucked too :(
Well, what I'm saying is : It's not too difficult for us to recognize such positions. But handpicking a large enough database might be tedious. An automatic filtering is desired but it's a step up compared to the previous case
(Probably doable, if you feed a decent enough classifier with a large enough labeled set but may take a while. Or maybe it's really easy and I'm overlooking something, haven't really thought about it in depth.)
