The End of the Golden Era of American Chess • Hacker News Board

Re: The End of the Golden Era of American Chess

#61

Post by DiogenesKynikos » Mon, Sep 16, 2019, 7:15 PM UTC

mindgam3 wrote: post #14

tus88 wrote: post #10

Earlier quoted context omitted.

How has computer chess changed how humans play humans? (I.e. when you take the computer out of thr equation).

At the amateur level, not much. Other than most amateurs spend way too much time relying on engine evals, and as a result don’t develop their own ability to learn “chess sense”. Competitive players can now prepare easily for individual opponents, which is cool and used to only be feasible for pros. For pros, it means that everyone plays the same safe openings, because they’re all “booked up” thanks to engines. Also m…

Interestingly, though, chess computers have also shown that even at superhuman skill levels, decisive, exciting games are possible.

AlphaZero and Lc0 have played many beautiful games against Stockfish. All three of these engines are far stronger than the strongest grandmasters. That implies that the drawishness of classical chess has to do with the conservatism of top grandmasters, who generally aren't willing to play the dynamic, seemingly risky style of an AlphaZero or Lc0. You have to really trust your own intuition and calculating abilities to play in that style.

Re: The End of the Golden Era of American Chess

#62

Post by mindgam3 » Mon, Sep 16, 2019, 7:26 PM UTC

Cpoll wrote: post #60

mindgam3 wrote: post #14

Earlier quoted context omitted.

At the amateur level, not much. Other than most amateurs spend way too much time relying on engine evals, and as a result don’t develop their own ability to learn “chess sense”. Competitive players can now prepare easily for individual opponents, which is cool and used to only be feasible for pros. For pros, it means that everyone plays the same safe openings, because they’re all “booked up” thanks to engines. Also m…

I hear a lot of claims that 960 eliminates opening theory, but is that really the case? I have no experience in the format, but I imagined that patterns would emerge (and Chess players are already good at spotting transpositions), and people would start studying a different form of opening theory.

Chess960 eliminates opening theory... for now.

I expect Chess960 to rise significantly in popularity and relevance as classical chess declines, at which point theory will become developed.

But that will take a few years, and even then, being booked up in 960 will only provide a small edge compared to classical.

Re: The End of the Golden Era of American Chess

#63

Post by mindgam3 » Mon, Sep 16, 2019, 7:35 PM UTC

DiogenesKynikos wrote: post #61

mindgam3 wrote: post #14

Earlier quoted context omitted.

At the amateur level, not much. Other than most amateurs spend way too much time relying on engine evals, and as a result don’t develop their own ability to learn “chess sense”. Competitive players can now prepare easily for individual opponents, which is cool and used to only be feasible for pros. For pros, it means that everyone plays the same safe openings, because they’re all “booked up” thanks to engines. Also m…

Interestingly, though, chess computers have also shown that even at superhuman skill levels, decisive, exciting games are possible. AlphaZero and Lc0 have played many beautiful games against Stockfish. All three of these engines are far stronger than the strongest grandmasters. That implies that the drawishness of classical chess has to do with the conservatism of top grandmasters, who generally aren't willing to pla…

I agree, it’s not all just due to engines. Carlsen is usually known for a more intuitive playing style, versus Caruana who is basically at the most extreme end of conservative calculator. If Lc0 can inspire the Caruanas of the world to play with more panache, there would be much rejoicing.

Re: The End of the Golden Era of American Chess

#64

Post by websitejanitor » Mon, Sep 16, 2019, 7:51 PM UTC

gridspy wrote: post #18

If (like me) you're bored by regular chess now, consider trying out the variant 16x16 "Four Kings one War." It has been turned into Steam game with a free demo : https://store.steampowered.com/app/883680/Four_Kings_One_War... The AI for this game was coded in Rust, by me. Good luck!

Also recommended for bored chess players, Shogi: https://en.wikipedia.org/wiki/Shogi

Re: The End of the Golden Era of American Chess

#65

Post by mindgam3 » Mon, Sep 16, 2019, 8:45 PM UTC

soulofmischief wrote: post #46

mindgam3 wrote: post #39

Earlier quoted context omitted.

Mainly full stack web (Node JS). I have a functional proof of concept integrated with Lichess and Stockfish API + custom chess code. It needs about week of love from a full stack dev to turn into an MVP I can test on actual humans.

Ok, let's talk. Email in bio.

Okay. Email sent.

Re: The End of the Golden Era of American Chess

#66

Post by shagie » Tue, Sep 17, 2019, 4:04 AM UTC

V-2 wrote: post #31

shagie wrote: post #28

Earlier quoted context omitted.

The game Arimaa attempted to do that. https://en.m.wikipedia.org/wiki/Arimaa

Is it more computationally complex than Go?

This game is played with chess pieces on a chess board... though they have different meanings.

Each player sets up their pieces in any way they want on the back two rows. The number of possible of possible starting positions for each player is 64,864,800. This means that the starting game position possibility is 4.2e15.

Next is the branching factor. Chess is about 35 possible moves each play for each player. Arimaa, each player makes two successive moves. This makes for a very large branching factor... 17,281 possible moves each turn ( http://arimaa.janzert.com/bf_study/ ). Go has a branching factor of 250.

https://en.wikipedia.org/wiki/Game_complexity has a chart on it... Go has a larger board size and space state complexity, but Arimaa has a larger game tree size and branching factor.

That 35 and 17,281 thing... 35^8 ~= 17,000^3. If you have a computer program that can search all the possible moves 8 deep in chess (Deep Blue went 6-8 typically), it could only look 3 deep into Arimaa (a similar search space is 5 deep in Go).