The Busted Dutch

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Ovyron
Posts: 2750
Joined: Tue Jul 03, 2007 2:30 am

Re: The Busted Dutch

Post by Ovyron » Tue Sep 24, 2019 4:49 pm

Uri Blass wrote:
Mon Sep 23, 2019 2:11 am
Note that close to perfection does not mean perfect so one game that you win does not prove it is not close to perfection.
What I was talking about was "close enough to perfection that you can't beat it."

Another way to express the same idea is "how much elo do you add on a take-back?", because analysis of chess positions is nothing more than playing variations until you find a winning line, or taking back the moves and trying another line until one of them is good enough to win.

The fatal flaw of Stockfish at any Depth is that after it's done searching it has to make a decision and make a move and there's no turning back, while one can check if a line is strong to defeat it, and if it isn't, try another one. Every line that makes progress against a given Depth would have a higher elo, so eventually you find a line that does win and that'll have a higher elo that than Stockfish's depth.

Stockfish at Depth 26 is already unbeatable with infinite takebacks, because there's always another line with a better score in a previous position so it'd take back its moves and play it instead and never get mated. Which means we could probably match the elo of Stockfish with X number of take-backs to the elo of Stockfish at some Depth Y, and then it can be proven that Stockfish Depth 70 isn't "close enough to perfection that you can't beat it" because there's some number of take-backs that go over its elo (eventually you can find a path where it plays a losing move, and you're done.)
Great spirits have always encountered violent opposition from mediocre minds.

Post Reply