Ideal Chess – drop chess perfected? | Hacker News Reader

Ideal Chess – drop chess perfected?

lesswrong.com

DeepMind and Kramnik had an interesting paper a few months back exploring possible chess variants:

https://arxiv.org/abs/2009.04374

The basic idea is to train AlphaZero on proposed variants, and look for ones that have fewer draws and a smaller advantage for white.

Interesting paper concept. I'll read it.

I also notice Kramnik is the last-listed author in a group of four, the rest of whom work at DeepMind. Is this a new marketing "thing" in published corporate research?

You may already be aware of this, but for those who don't: depending on the discipline and other factors, the last author in the list may be known as the "anchor". Often, frankly, the preceding authors do the majority of the work, while the anchor functions as a supervisor.

I'm ignoring "vanity" authorship and other horrors.

Did you read the paper? It is quite long, mainly because of Kramnik's manual analysis of some selected games. It appears his role was to explore game playing factors, which are hard to catch with statistics.

I'm excited to see this. It reminds me of a paper by Cameron Browne (http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/03/browne_...), which does similarly, but using less advanced technology. While the approach was janky in some ways, the game that the system invented, Yavalath, has received some praise for being an interesting and compelling take on the n-in-a-row genre.

I've been dreaming of doing similar with an alpha-type system, and so I look forward to reading this paper.

Edit: After skimming it, it doesn't look like they're going the distance like Browne did. Browne had the computer evolve games that the AI would rate ideally, while the DeepMind paper only has the computer consider a handful of pre-selected variants. I don't anticipate it would have been much extra work to evolve games, and I would have found it much more interesting.

Here's an idea (probably not original): simultaneous chess. Both players write down their moves and, once they have both committed, both moves are executed in parallel. If two moving pieces end up on the same square, both are eliminated. If one piece moves onto a square that has been vacated by an opponent's piece in the same move then no capture takes place.

Addendum: One benefit of this variant is that it is completely symmetric and hence completely fair. There is no advantage for either side because neither side moves first.

KungfuChess online multiplayer

Not the same thing. KFC is real-time.

Fun idea, but what happens when a player submits an illegal move that becomes legal only when taking the opponents move into account or that’s only legal before the opponent’s move? Say a rook moving from a1 to a3 while an opponent’s knight moves from or into a2?

Also, if you are going to write down moves, you can allow the players to move all of their pieces, too, like in Diplomacy.

A queen could capture a pawn next to the king, and a bishop could simultaneously move to prevent the king from capturing the queen, for example. A countermove could be to move a pawn between that bishop and that queen, and to have the king capture the queen.

⇒ I guess this multi-move variant would get highly complicated very soon. It might be better to follow diplomacy here and replace “If two moving pieces end up on the same square, both are eliminated” by “...same square, neither moves”, or maybe add “x supports y moving to square S”*?

I think I would specify that a move has to be legal when at the start of the move, i.e. while it is being written down.

Or alternatively, the piece doesn't move if the move turns out to be illegal? Would have to try both options to see which works better

That would open up the possibility of a paradoxical situation where both moves are legal at the start but neither is legal at the end.

So, a king could move into a checkmate, even if it wasn’t in check at the start of the turn. Seems fine to me, but I think it would be funnier if a king in check could stay put, with the player in check betting that the opponent’s next move will end the check.

There also is the possibility of simultaneous checkmate.

Another edge case is pawns taking pieces. Could a pawn move diagonally into a square being vacated by the piece standing there? Could two pawns simultaneously not take out each other (say by the simultaneous moves d4×e5 and e5×d4)?

> a king in check could stay put, with the player in check betting that the opponent’s next move will end the check.

Yes, that was part of the intent. There is game theory in play as well as chess strategy.

> There also is the possibility of simultaneous checkmate.

Yes, or even simultaneous king capture without checkmate(s) resulting in a draw.

Or you could adopt a quiddich-like rule and say that if the kings are simultaneously captured then the game is decided by the sum total of the remaining value of the pieces on each side.

> Could a pawn move diagonally into a square being vacated by the piece standing there? Could two pawns simultaneously not take out each other (say by the simultaneous moves d4×e5 and e5×d4)?

I would say yes and yes.

I wonder how such a game would end.

With no way to block or capture a piece on the square it's attacking from, the first time the king is placed in check he has to move or be captured. If the king isn't completely surrounded, the attacker only wins if they successfully guess where (or if) the king will move. The endgame would include an element of reading the other player, like poker.

> The endgame would include an element of reading the other player, like poker.

Yep!

Note that it is also possible for the game to end in a draw if both kings are captured on the same move.

You should try Diplomacy :)

That's where the idea came from :-) But I played Diplomacy almost 40 years ago (God help me), I wasn't sure if it was still a thing.

Definitely still a thing! I've picked up recently, there's a couple of websites where you can play for free. I've played in playdiplomacy.com.

I had an idea for a chess variant recently that I haven't had time to test yet (but maybe it already exists?): it's exactly the same as normal chess, except on your turn you can move any piece, black or white, even your opponent's king (God chess?) You still have to eliminate the other king, and a piece can still only take pieces of the opposed color, obviously. I suppose it very well might turn out to be completely awful, but in my head it sounds interesting.

My idea for a chess variant was giving every piece a fixed number of moves (with move-count visible at all times). After the quota of moves is finished, the piece becomes immobile except that it can still capture an enemy piece (and thus move to the captured square).

Sounds interesting! It could be played easily on a physical board by putting poker chips under the pieces. Without a way to replenish counts there is a risk of deadlock, though. Maybe an option to pass to recover one move on a piece of your choice?

I hadn't thought that far! The poker chips are a great idea. Perhaps a reset mechanism after both sides run out of moves would work too.

> but in my head it sounds interesting.

It really does sound interesting to me as well. I suggest you grab a friend or random chess-playing stranger and challenge them to a game of "God-mode chess".

One jerk move possible - just undo the other player's move the moment that becomes possible (pawns can't move backwards, for example).

This seems like it'd degenerate into "pawn moves only", since you can immediately undo any other move.

Captures cannot be undone either, but yes, that might be a problem. Go solves it with the ko/superko rule that states it is illegal to move in such a way to reproduce a previous configuration of the board, and I think the same rule would be relevant here.

Edit: another simple solution (with interesting strategic implications) could be to forbid moving the same piece twice in a row.

The best way to solve this is to say that you can't move a piece right back to where it just came from, like the ko rule in Go. As for superko, I think the threefold rule works just fine.

Somehow this reminded me of a football variant (soccer for americans) that has two balls in play, either ball is good as the other.

I wondered about this recently and assumed one reason this sort of thing doesn't catch on for mainstream sports like football (soccer) or basketball is because it'd be a mess to spectate/broadcast. Sounds fun though!

Incidentally there seems to be a chess/draughts like game rather inspired by football, called Diabolik.

https://en.wikipedia.org/wiki/Diaballik

In university, friends in the lab came up with, and regularly played, a 4-player version we called "Drop Chess".

In this variant, there are two games going on at once asynchronously. Your partner is sitting beside you, but playing the opposite colour. Whatever pieces he captures, you can drop on the board- but not until the next turn after he captured it, and with some extra rules about where it can go. When one game ends, both games are over.

It was as much fun to watch as play.

This was a regular thing at tournaments I used to play as a kid. We called it “transfer”.

You know, come to think of it, probably someone in the lab learned it elsewhere and brought it.

None of us were bright enough to come up with that gem.

That’s called bughouse chess and it is quite popular.

Here's a variant my 8 year-old niece invented, and it has some great playability:

normal rules, except each side makes two turns in a row.

Constraint: the second move cannot logically depend on the first move in any way.

We've played a few dozen games this way. Really opens up some interesting possibilities.

That sounds so fun. Honestly, I wonder how you could enforce that on more experienced chess players; is there a way to make a second move "forgetting" what you did first?

"Logically depend" has an easy interpretation: the second move must be also be valid if performed before the first move. [You could do a harder variant where both moves must be legal simultaneously.]

I assume it just means that, for instance, you can't move the same piece twice, or for the second move to one only legal once the first occurs.

This is already a known variant (modulo the constraint on the second move not relying on the first). Best way to play it is that the white's first turn is only one move, then black starts with the first double move - this is the 12* protocol (first implemented in a game called Connect6), and it practically eliminates any first move advantage

A comment in the original article says it best: ... Changing the rules tends to neutralize acquired knowledge ... and beginning players like this because it levels the playing field.

^This, but its a good thing. Something that irks me about Chess is that at the high level the conventional wisdom is to memorize openings and gambits rather than derive moves from some sort of first principals. IMO this undermines the entire point of a strategy game.

On the other hand the best player in the world is known to be a player who puts less effort into opening prep and takes his opponents out of book very frequently.

Fischer Random Chess (aka Chess960) is meant to address exactly that: https://en.wikipedia.org/wiki/Fischer_random_chess

It has a large online following.

It doesn't seem to be that large of an online following: it's supported on the main chess sites but not as one of the top default modes.

And yet the inventors of Capablanca Chess and Fischer960 chess were both World Champions

I’ve been wondering if there is a way to change chess so that AI is no longer the obvious winner. Chess itself went through a variety of rule changes throughout its history, so this wouldn’t be completely novel.

Off the top of my head - somehow creating new rules, especially socially-dependent or open-ended ones, on demand might work.

I asked a similar question a few months ago and got some interesting answers:

https://news.ycombinator.com/item?id=24441073

Otherwise, I really liked this chess variant, although it seems near impossible to actually play:

https://en.wikipedia.org/wiki/Alice_Chess

Alice chess is a chess variant invented in 1953 by V. R. Parton which employs two chessboards rather than one,[a] and a slight (but significant) alteration to the standard rules of chess. The game is named after the main character "Alice" in Lewis Carroll's work Through the Looking-Glass, where transport through the mirror into an alternate world is portrayed on the chessboards by the after-move transfer of chess pieces between boards A and B.

And there's chess boxing, which combines chess and the sport of boxing. Robots in theory could be designed to play it well, but humans definitely have the chess boxing edge over server racks.

https://en.wikipedia.org/wiki/Chess_boxing

Arimaa is a game that was designed to be hard for AI to master. Its rules are quite different from chess, but it has a lot of chess in its DNA. It has since been solved, but it survived 11 years with a $10,000+ prize for the first computer to beat a human champion.

Now that we are in the age of AlphaZero, I think we are close to the bottom of the barrel of ways to design games to hinder AI. I would be surprised if I saw another pure abstract strategy game that favored humans over AI these days, the way Go and Arimaa used to.

This is interesting, thanks for the suggestion.

Promotions are a big part of shogi, and part of what makes it fun. You do not start with the most powerful pieces, you have to earn them through promotions. Also, the promotion area (the enemy camp) is larger. This article doesn't capture that aspect of the game... which is necessary if you are looking for an ideal chess variant.

You may also want to take a look at some mechanics of Xiangqi, another chess variant.

I agree that promotions are an important part of what makes Shogi work so well. As I mentioned in the article, I see there are two ways to make drop chess work - either build a game around weak pieces, or make the king more nimble. I think Shogi has perfected the first approach. I love Shogi. I don't want to reinvent Shogi. But there's something that I love about Western Chess: I love knights, bishops, rooks, and queens. I also like the pawn structures it has. I like Western Chess, but I don't love it the way I love Go or Shogi. My goal is to create a version of chess that I love, and has the knights, bishops, rooks, and pawn structures that I love. This is my current best guess at what that looks like.

I do not like that pawns move forward but capture diagonally, or en-passant. I also do not like the teleportation that takes place during castling. I do not like that many pieces can cross the entire board (bishops, rooks, queens), and that they're so many of them.

I think shogi already took care of many of the things I do not like about western chess.

If you want to have fun playing with chess variants, check out Really Bad Chess[0], it's a random variation each day that you play against AI and rank yourself against other players.

[0] http://reallybadchess.com/

I played a good bit of Really Bad Chess when I was first getting into chess. I'm a big fan of Zach Gage's games, I think compared to his other works it's not as platonically good as his other works, but what I love about RBC is that the difficulty of the board is chosen based on your skill level, so when you're a beginner you get positions that are absurdly unbalanced in your favor, but then as you get better you eventually get the tables turned and have to work your way out of a difficult situation. It goes a long way to make chess feel more accessible and enjoyable for someone who doesn't know what they're doing.

That said, Really Bad Chess isn't really a chess variant. It's just orthodox chess with a semi-randomized board layout, but it changes none of the core mechanics, and all the pieces are the same as orthodox chess

Perhaps a variant can be devised which causes a dice roll at the time of each capture. The attacker needs to roll more than the number of direct supports a piece is having to take it, otherwise they lose the attacking piece.

It will be a bit like XCOM, before a move one will know the percentage of success of various captures, and will have to make a decision in that light.

So even a higly probable capture has a small chance of failure, sometimes resulting in the player in a weaker position being able to stage a spectacular comeback victory. In normal chess it perhaps impossible to win from a disadvantaged position.

Is there any such chess varient which uses dice, to add a chance element?

This reminds me of David Sirlin's ill-named 'Chess 2' - one notable mechanic from there is that players have stones that they can use to bid when a piece is captured, and if the defender wins the bid, then the capture fails.

My idea for a chess variant is that an engine autoplays the first X moves into a sharp, balanced position and this is what is used by all tournament players for the round.

This would avoid excessive opening preparation and steering opening to drawish positions.

An engine would do the opposite of that. Chess engine competitions need to use human-prepared, unbalanced opening positions because otherwise it would be a draw fest.

If you've abandoned chess.com for lichess.org, you might be surprised to learn that chess.com now hosts 15 variants.

Is there any variant where both players play blind every other move or in the starting?

There's a 'fog of war' variant where you can only see the squares that a piece of yours could move to.

--that a piece of yours could attack, I believe, not move to

I just checked and if you want to be technical it's "move to normally, could move to during a capture if a piece was there, or holds a piece you could capture".

Pawns are the only case that where those three are different but they see all three:

- directly in front (move)

- forward-diagonal (normal attack)

- horizonal exactly only in the case of a enemy pawn moved to that square last turn that can be taken en-passant (capture). That square goes dark if you move another piece.

That one would be interesting to try with AlphaZero. Reinforcement learning systems tend to have trouble with imperfect information.

pychess.org has many chess variants, and you can play against the AI to try them out. No registration required.

Some time ago I also had the idea of procedurally generated games, not only of chess, but using basically all the known rule sets and pieces of all known games.

I captured a lot of game details and have them stored away, but I had to move onto something else shortly after, so I never explored it further.

If you haven't read this 2008 paper on procedurally generated abstract strategy games, I'd say it's related to your interests: http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/03/browne_...

The resulting game, Yavalath, is well-known in the abstract strategy game community: https://boardgamegeek.com/blogpost/2814/yavalath-evolutionar...

So an entire blog post on a game he hasn't tried yet? That's in the spirit of cooking blog posts where the author has cooked the dish once.

I'm reminded of most proposals to defang Lisp of its parentheses.

In my defense, I originally published this as a throwaway on my Shortform (it's sort of like LessWrong's take on Twitter), but one of the admins wanted me to repost it as a main post. I agree that ideally I would have wanted to playtest this more before drawing attention to it.

That being said, I do suspect that these rules are probably pretty close to achieving the effect I wanted to achieve- the main doubt I have is whether the king's moves work, or if they need to be adjusted to get the balance right, but even there, there's a good chance that it works. I will make sure to update the article when I get a chance to test it out.

I think a sort of reversed drop chess would be interesting: start with just a king each, and each turn you can either make a normal move or place one of your fifteen other pieces on its starting square.

Go baduk weiqi is still worth looking into for chess players.

Author of the linked post here. I've actually done more Go than Chess in the past 6 months - I enjoy it a lot. My dream is to find a game that captures what I love about both Go and Chess. I've seen many examples of how not to make that game, but my mind turns back to that question every now and then.

The lack of rules while opening up the world of complexity is what makes go interesting. Since chess involves 'functional pieces' I feel it may be doomed from the start. The two games may be scratching entirely different itches I'm afraid.

Allow to move a piece in go

Fisher invented https://en.wikipedia.org/wiki/Fischer_random_chess

Reply on news.ycombinator.com