Computers are getting better and better at evaluing how strong a position is, and what the next move should be.
Sometimes, it puzzles us and we mere humans have some difficulties trying to grasp the logic behind such or such move found by a computer.
So, is there somewhere someone who tried to evaluate the difficulty of a position ? I know lichess does it for its puzzles, but I guess it's just an average ELO of people who found the solution of the puzzle.
It's quite tricky I think, as the concept of "difficulty" involves more concept : knowledge of the player (like the first moves of an opening or the last move of a finale), obviousness of the good move, etc.
What do you think ?
To give an example, for a computer, King+Queen versus King is a 100% win, as well as King+Bishop+Knight vs King. The two positions are as good as the other, because they both lead to victory.
However, they are certainly not the same difficulty.
So a difficult one will be the position where you have a lot of possible moves, and a very small amount of good moves. All the tactically sharp positions become complex. A complex position will be the one where you can't find good moves in an obvious way. The kind of positions with big imbalances where evaluation at different depths with the engine varies a lot. The B+N is much more a technical thing once you know how to do it...
Well, there are mates beyond 500 moves in the 7-men-TB.
I find them not that easy! ;)
A similar approach to the ELO option would be to, say, compare AI's older versions' outputs with newer ones.
"Lot to lose with but few moves" also shortsells long-term positional cracking; the grinding MC demonstrates.
@Sarg0n wasn't there some sort of issue that didn't display the shortest way to mate directly available? I mean 500 is gross. Can you tell what pieces are involved? (my guess is some Q vs B+B+N+N) Also there is a very famous and odd ending involving the h or a pawn for one side and two bare knights for the other which is theoreatically possible to win.
@lement Magnuses cracking is much more about mindgames and his opponents overassesing his chances with slight advantages which they don't like to concede so they avoid concrete lines, which makes them concede bigger advantages which they also don't want to fix, etc. For instance if you take Nepo's style and his score against Magnus you'll see he doesn't do well against harsh oposition and concrete lines all the time.
Computers cannot adequately evaluate the difficulty of a position from a human perspective. If they did then they would be able to mimic a human player then it would become almost impossible to detect a cheater and that would be the end of online chess.
Because difficulty is directly related to 'how' you find the right move. And computers and human brains work based on totally different mechanisms and concepts.
@Truth_Teller wouldn't be the end totally.. Us humans will always adapt and overcome.. eventually.... to get on subject of detecting difficulty ... ChessTempo classifies their tactics like removing the defender... yada yada . lets say players below like 1500 struggle with that concept for them it would be hard.. so lets give that type of puzzle a score of 1 or easy.. then have a similiar puzzle thats like removing the defender but introduce a combo to set up that tactical position and suddenly the puzzle gets harder ... its not an eloquent solution but would need classifying every tactical position by type and just have a base ... I don't even know if im coming across coherent lol
