Depth of one is enough • page 2/2 • General Chess Discussion • lichess.org

Thanks for the link. @kalafiorczyk
I started a simple test run of ways to sort the games from fewer mistakes to more of them.
It's a work in progress, for the next few week. Not sure if my study will have the full a.tse file done, but it's a hobby and an eye opener at the same time for me.
https://lichess.org/study/Vhvyt9XE/

Thanks for the link. @kalafiorczyk I started a simple test run of ways to sort the games from fewer mistakes to more of them. It's a work in progress, for the next few week. Not sure if my study will have the full a.tse file done, but it's a hobby and an eye opener at the same time for me. https://lichess.org/study/Vhvyt9XE/

kalafiorczyk

#12

@Toscani said in #11:

I started a simple test run of ways to sort the games from fewer mistakes to more of them.
Thanks again for the explanation. I don't know about Stockfish, but some years ago people did run similar experiments using the then-champion Rybka engine. The strength of Rybka with single valid move evaluation (more or less equivalent to "go nodes 1", purely based on move ordering) was estimated in range 1200-1500. And Stockfish implements multiple enhancements over relatively ancient Rybka. Obviously, they both play horribly bad and stupid endgames, but still are noticeably competent in the opening and middle-game.

I don't wanna discourage your from pursuing your experiments, but maybe try to search for some old treads in relation to weakening Rybka in human-like fashion. Stockfish is well-known to be making inhuman moves even with "UCI_LimitStrength" enabled and low depth. That is how they keep confidence that the cheater-catching algorithms are still working, even for the relatively weak players.

@Toscani said in #11: > I started a simple test run of ways to sort the games from fewer mistakes to more of them. Thanks again for the explanation. I don't know about Stockfish, but some years ago people did run similar experiments using the then-champion Rybka engine. The strength of Rybka with single valid move evaluation (more or less equivalent to "go nodes 1", purely based on move ordering) was estimated in range 1200-1500. And Stockfish implements multiple enhancements over relatively ancient Rybka. Obviously, they both play horribly bad and stupid endgames, but still are noticeably competent in the opening and middle-game. I don't wanna discourage your from pursuing your experiments, but maybe try to search for some old treads in relation to weakening Rybka in human-like fashion. Stockfish is well-known to be making inhuman moves even with "UCI_LimitStrength" enabled and low depth. That is how they keep confidence that the cheater-catching algorithms are still working, even for the relatively weak players.