lichess.org

Timeout in a dead position

Shouldn't this count as a draw even if time run out as mate can't be reached by any series of possible moves?

lichess.org/a1GMMavC


DRAWN. Checkmate isn't possible through any sequence of legal moves, despite their being sufficient mating material on the board.

there neeeds to be a terabyte of information cuz there are hundreds of positions like this

In theory an " dead-position detector" may be feasible. Ideally such positions should be terminated at the spot and decleared draw.

I could code up something that checks this, but it would be insanely computationally expensive, and wouldn't work outside of pawn locked, king locked positions like this (and the worrying thing is false positives).

„at the spot“ LOL

Do you have a mega-cluster at hand to calculate all timed-out positions? Thouroughly and properly? Instantly to depth 100?

They have tried to code it for years and decades. Fortress recognition is a tough nut to crack.

Should ask google to create OmegaZero...

A dead position recognizer could be coded. Do something like this:

Player 1 times out and loses.
Go to the "computer analysis" board and request computer analysis.
Computer attempts to calculate to mate using "Bad stock" (tries to play worst move) for the side that ran out, and "Stock 8" for the side that did not time out. Giving the timed out side the worst possible moves, and the best possible for the side that had time. If 50 moves go on without a checkmate the position of bad stock vs stock it should be declared a "dead draw" and the game result should be revised to be a draw.

If it sounds easy, read this. See if you or your engines can tell the difference in the two positions at the end.

chess.stackexchange.com/a/22764/15989

Also read the other answers and their comments to see why some intuitive heuristics don't work.

Reconnecting