drj4759 wrote:There is no exact boundary of opening, middlegame and endgame, my statement is a generalization. Chess endgame definition is dependent on the person who interprets its, it means so many parameters.
Current chess engines technology could not solve opening, middle game or endgame by itself. It is the opening database or endgame database that is making it happen as these are reducing it to science. Middlegame database will wait until the opening database is near perfect.
For as long as there are idiot chess engines, Stockfish and Komodo will continue to gain ELO points. If Stockfish, Brainfish or Komodo will self-play against themselves, the ELO difference is close to 0 even without opening or endgame databases. If they make improvements from their previous versions, that is where the ELO is gained.
Chess engines ELO ratings are not standard. One can choose what they want. I can have top ELO rating of 10,000 in my rating list if I am inspired. Right now, Stockfish 8 is 3500+ in my rating list.
I consider the chess game solved when Stockfish vs. Komodo or any top chess engine play against each other with the help of opening + middlegame + endgame books and it always ends in a draw. It is not yet here today, but it will be probably in the future.
can't you really understand it?
what Cerebellum would do would do is completely meaningless. You say it will analyse to ply 36, and then analyse another 36 plies deeper, etc. Imagine the following scenario: at ply 3 SF has a node with eval 30cps, which is actually just 10cps, and another node with eval -20cps, which is actually 40cps. It prunes the second node in some way, and further proceeds to depth 36 with the first node, and then another 36 plies, returning the score. Now, that 'best move' is completely meaningless, as SF has already missed the much better other choice because of wrong eval.
Until engines have perfect eval, and try to prune even a single node, which they do amply, finding an ultimate best move is simply impossible.