Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Human Chess is a chess variant where playing the top engine move is forbidden

humanchess.abcd.party

Post reply on HN
Print view View original

71–80 of 243 posts

Re: Human Chess is a chess variant where playing the top engine move is forbidden

#71

Post by fernandopj »

BurningFrog wrote: post #68

I've been thinking about a possible chess variant to eliminate opening preparation drudgery. The Fisher 960 variant tries to do this, but it can be very different from regular chess, and some of the positions are unbalanced. I think we can use the fact that engines know when a position is even. There must be millions of even positions in the first 10 or so moves. Pick one of those randomly, and start the game.

This is already how some engines tournaments work. They don't start from move zero, they start from some uncommon position after a few moves, but one still considered even or at least not unbalanced.

Re: Human Chess is a chess variant where playing the top engine move is forbidden

#72

Post by trsohmers »

tda wrote: post #35
lesiki wrote: post #15

Now: Human Chess: a variant where you can't play what AI would play. Next: AI that can play Human Chess. After: Human^2 Chess: you can't play what the AI above would play. etc I wonder if this creates distinctly new games at each level, or if it's just nonsense one level down.

rinse and repeat that a few times, and the only remaining winning move will be not to play

Would you like to play a nice game of Global Thermonuclear War?

Re: Human Chess is a chess variant where playing the top engine move is forbidden

#74

Post by kadoban »

rocqua wrote: post #58
kadoban wrote: post #38

Earlier quoted context omitted.

Yeah, deterministic is not actually the concept I want here I think. It's fine if it's random, it just needs to be unambiguous. So you just choose in advance what settings to run with and the stopping condition. And then it doesn't matter that if you had run it with different settings, you may have gotten a different answer.

It is less fun if the outcome is non-deterministic. It means that occasionally the win is determined randomly. That takes away a certain element of skill.

I guess if you let it run for a long time it should converge on a first move?

I don't think there's a real fix for the issue, unless someone effectively solves chess someday. Otherwise your win/loss is fundamentally based on the imperfect evaluation of a particular engine.

If it's really just the nondeterminism that bothers you (which is fair enough, preferences vary), there's engines that either are deterministic or can be made so with settings.

Re: Human Chess is a chess variant where playing the top engine move is forbidden

#75

Post by ccozan »

vanderZwan wrote: post #39
mattmaroon wrote: post #33

Earlier quoted context omitted.

Unless it was two different engines.

Are we reinventing GANs for chess engines here, or does it just happen to sound kinda similar?

I'd say let's go straight to SD and let AI paint the next move!

Re: Human Chess is a chess variant where playing the top engine move is forbidden

#77

Post by xeyownt »

vanderZwan wrote: post #52
shkkmo wrote: post #45

This seems like a misnomer of a name. "Human Chess", a variant of chess that can only be played if you have a computer...

I think it is a joke on the accusations among professional chess players that some players cheat by using a chess engine to determine their next move. In other words: claiming the move of your opponent is the top engine move is equivalent to accusing them of "cheating".

Yes, but this new rule also adds interesting new mechanisms, like for instance the kamikaze check move with the queen, where the opponent's only move is to take with the king, hence losing the game (as it is also engine top move).

Mastering that kind of new threats does not seem easy IMO, and in fact could well be mastered by... computers ;-)

Re: Human Chess is a chess variant where playing the top engine move is forbidden

#78

Post by jetnew »

If we train recursively-restricted reinforcement learning agents, could there be interesting differences in the behaviors that emerge? Could it even be used as a method for exploration?

Some set-up considerations: 1) Actions must be discrete, or at least binned for restriction, 2) The number of times to restrict is limited by the size of the action space

I would imagine for CartPole, the balancing would become more wobbly, while still somewhat successfully balancing. But in more complicated environments, it could result in much more different behaviors because the states visited (and trajectories) could be different.

Re: Human Chess is a chess variant where playing the top engine move is forbidden

#79

Post by tomesco »

sebstefan wrote: post #49

It's a cool concept but the first naive thought that comes to mind for me is that white could just easily dominate by taking advantage of the fact that taking a queen is almost always the top engine move. So just by maintaining relative king safety on white's side, you just open your queen early and make sure that every move there onwards hangs your queen in some fashion. It's very easy to hang a queen. You just have…

There must be a threshold above which making an engine move is allowed. Because if there's a checkmate move, it will be the best engine move and needs to be allowed.
Post reply on HN