Definition of a blunder

[Uri Blass]

For definition of a blunder I thought to define it based on the change in expected result.
Something like if the change in expected result is at least 0.2.

The main problem is how to calculate expected result.
We need a good function that tell us the expected result for every position.
The simplest way is to translate stockfish's evaluation to expected result but I think that it is not a good idea because stockfish like other engines is stupid and does not understand that the opponent may not see something and that the expected result for mate in 8 that the human opponent may miss is not 100% loss and this mate in 8 may be better than being a queen down without practical chances even if the engine does not see a mate score.

I also think that stockfish sometimes is also too smart and see at small depths big evaluations that humans do not understand and because they do not understand them they are not correct from practical point of humans.

I think that we need special chess engines to evaluate expected result in every position when they get only position and the rating of the human players.

The engines need to give expected result of every position in the game before seeing the rest of the game and their target is to minimize the error that can be calculated as (expected minus observed)^2

We can make a competition between engines if there are some engines that do it when we simply calculate for every engine the sum of errors it get and the engine who get the smallest error win.

The question is if there are programmers who are interested to make a special engine for this competition when the engine does not need to return a move but only to return expected result for a position.