Chess is fairly well known, but there's also an entire world of chess variants, games that take the core ideas of chess and change either a few details or completely reimagine the game, either to improve the game, or just change the flavour of the game. There's even an entire website dedicated to documenting different variants of chess.

Today I want to tell you about some classic chess variants: Crazyhouse chess, Grand chess, and Shogi (Japanese chess), and posit a combination of the first two that I suspect may become my favorite chess when I have a chance to try it.


Shogi is the version of chess that is native to Japan, and it is wildly different from western chess - both western chess and shogi have evolved continuously from the original chaturanga as the game spread out from India. The core difference between shogi and the familiar western chess is that once a piece has been captured, the capturing player may later place the piece back on the board as his own piece. But if that was the only difference, it would make for a very crazy game, since the pieces in western chess are so powerful while the king is so weak, that the game would be filled with precarious situations that would require the players to always have their guard up for an unexpected piece drop, and checkmate is never more than a few moves away unless both players are paying close attention.

In fact, this version is precisely crazyhouse chess, and this property is both what makes crazyhouse chess so beloved and fun, but also what stands in its way of being taken as seriously as orthodox chess. There are two ways that this barrier could be overcome - either the king can be buffed, giving him more mobility to better dodge the insanity that the drops create, or the pieces can be nerfed, making them much less powerful, and particularly to have less influence at a long range. Shogi chooses the route of nerfing the pieces, replacing the long-ranged and very influential pieces used in orthodox chess with a set of pieces that have much more limited mobility, such as the lance, which moves like a rook, but can only move straight forward (thereby limiting its position to a single track), the uma (horse), which moves like a knight, but can only move in the two forwards most positions, or the gold and silver generals, who can only move in a subset of the directions that a king can move in. Since each piece isn't much stronger than a king, it is much easier for the king to dodge the threats produced by each piece, and a king can only be checkmated when the pieces are acting in coordination to create a trap for the king. (This is the basis of tsume-shogi, checkmate puzzles for shogi. They are fun to solve, and I recommend trying them out to get a feel for how different checkmates in shogi are from orthodox chess checkmates)

I think shogi and crazyhouse solve a problem that I have with modern orthodox chess: the game ends in draws far too often, and the endgame is just too sparse for my taste. You can get good puzzles out of orthodox endgames, but I find the endgames of shogi and crazyhouse to be much more fun and much more exciting.

I a.  (An aside)

While I'm on the topic of shogi and crazyhouse, shogi pieces look quite different from the pieces used in orthodox chess:

I quite like the look of these pieces, and it provides a solution to a practical problem that arises from the piece drop mechanic: With orthodox chess pieces, one would need two sets of chess pieces, a double-sized army for each player, since each player may have up to twice the regular amount of each type of piece after they capture the enemy’s pieces. With these flat, wedge shaped pieces, though, a player can just make the piece face in the opposite direction towards their opponent, and a single set of pieces is enough to play the game. While I think this solution works, and these pieces are quite iconic for shogi, it just doesn’t feel right to play crazyhouse chess with pieces like this: crazyhouse chess is orthodox chess at its core, and it feels right to play crazyhouse with orthodox chess pieces. My ideal solution would be pieces that are as tall as orthodox chess pieces, and have a similar design language, but which are anti-symmetric: the pieces would have flat tops and bottoms, and can be flipped upside down to change the colour of the piece, since one end would be white, and the other end would be black. I imagine the two colours would meet in the middle, with a diagonal slant so that it would show one colour primarily to one player, and the other colour to the other.


It's been an observation made more than once, that there's a certain feeling of completeness to the orthodox chess pieces: The rook and bishop each move straight in certain directions, either perpendicularly / parallel to the line of battle, or diagonally to the line of battle. If you were to draw a 5x5 square around each piece, the knight can move to precisely the squares that a rook and bishop can't go to. And the queen can be viewed as the sum of a rook and a bishop. It all feels very interconnected, and almost perfectly complete and platonic. Almost perfectly, because there's two sums that we don't have in orthodox chess: the combination of a rook + knight, and a bishop + knight. These pieces, called the marshal and cardinal are quite fun pieces to play with, and I would not argue that chess is a better game for omitting these pieces. As such, there have been proposals to add these pieces to the game, the most well-known of which are Capablanca chess and grand chess, proposed by Chess World Champion J. R. Capablanca and the game designer Christian Freeling, respectively. The main difference between the two is that Capablanca chess is played on a board 10 wide by 8 tall, while grand chess is played on a 10x10 board, with an empty file behind each player's pieces, aside from the rooks, which are placed in the very back corners (what about castling? Simple, you can't castle in grand chess):

The additional width of the board in Capablanca and grand chess is used to allow one each of the marshal and cardinal to be placed in each player's army. Aside from the additional pieces and larger board, grand chess plays just like regular chess, but I think it deserves to be considered seriously as an alternative to the traditional rules for chess.


While an introduction to chess variants would make a good topic for a post here, that's not what I'm writing right now. While these three games would certainly be present in such an article, the selection would be far too limited, and far too conservative - there's some really crazy, wacky, fun, and brilliant ideas in the world of chess variants which I won't be touching on today. I'm writing today because I want to talk about what I think may be the best contender as a replacement for orthodox chess, a cross between grand chess and crazyhouse, with a slight modification to better handle the drop mechanic of crazyhouse. It's clear that Capablanca chess and grand chess were intended from the very start as rivals to the standard ruleset, and I mentioned previously that shogi solves a problem that I have with orthodox chess: orthodox ends in too many draws, and I find orthodox endgames to be less exciting than crazyhouse and shogi endgames. My ideal game of chess would look more like crazyhouse than orthodox chess, since drops just make chess more fun. As I mentioned before, while crazyhouse is a fun game, it's just too intense and unpredictable to present a serious challenge to orthodox chess (at least, that is what I suspected as I was thinking about this post). There are two ways this can be addressed: the first is to do as shogi did, and make the pieces almost all as weak as the king, so the king can more easily survive against the enemy pieces; but doing this makes the game a different game from orthodox chess; it's no longer just a variant on orthodox chess, it's a completely different flavour. A flavour that I happen to love, but not the flavour of orthodox chess. I wanted a game that would preserve the heart of orthodox chess, while giving it the dynamic aspect allowed by drops, but more balanced and sane than crazyhouse chess.

So let's explore the second way to balance crazyhouse chess: instead of nerfing the pieces, let's make the king more formidable, more nimble, and able to more easily survive the intensity of drop chess. I haven't playtested this yet, but it seems appropriate to give the king the 4 backwards moves of the knight: This will give mobility to the king, without giving it too much mobility, and limiting the king to the backwards moves will ensure that it remains a defensive piece, and doesn't gain a new life as an aggressive part of the attacking force. Playtesting may prove this to be too weak (I don't anticipate that it will make it too strong): If this is the case, a different profile of movement may make sense for the king, but in any case, it is clear that increasing the mobility of the king will allow for a balanced form of drop chess.

Ideal Chess

So my ideal chess would differ from orthodox in the following ways:

  • The game is played on a 10x10 board, instead of the traditional 8x8 board (I feel that a wider board will make for a more fun, and deeper, game of chess)
  • The game will feature the marshall (rook + knight) and cardinal (bishop + knight) of grand chess, and will have the pieces arranged in the same way as grand chess (this also implies no castling)
  • When a piece is captured, it may be dropped back in to the game by the capturing player (working exactly as in crazyhouse chess or shogi)
  • The king may, in addition to its usual move, move using one of the 4 backwards moves of the knight. Pieces may be captured using this backwards move.

Ideally, the game would be played using the tall, bichromatic, antisymmetric pieces I propose in section I a of this post.

New Comment
23 comments, sorted by Click to highlight new comments since: Today at 11:56 PM

Excellent post!

I would greatly appreciate a follow-up, perhaps compilations of game variants for other popular games such as Go or Poker?

I'm not familiar with Poker, so I wouldn't have much to say there.

The way I see it, there isn't really such a thing as a Go variant, the same way there are Chess variants - there is an entire genre of Abstract Strategy Games (which chess also belongs to), of which placement games are a subset. Some notable placement games include Hex, Havannah, Six-in-a-row, Catchup, Blooms, and probably some good ones I'm forgetting right now. I suppose that among placement games, there are some that extend Go's rules directly: Sygo, Gonnect, Two-stone Go, Omino Go; but the same way I view chess as belonging to the chess variants category, I view Go as representing placement games, rather than Go variants.

I may on a whim write a post about abstract strategy games, but if I get around to that, I also want to write a more proper introduction to chess variants; as I mention in the main post, the list here is quite incomplete and conservative, serving merely to provide background to my own ideas that I discussed


Thank you for the response, I will definitely check out these variants. I'm trying to understand what sort of simple rules let deeply strategic games emerge out of them, and how inventors of such games come up with these ideas.

If you're curious to read about abstact game design, I recommend reading Christian Freeling's How I Invented Games and Why Not and perusing Nick Bentley's blog

Setting aside my intrinsic love for board games, the aspect of this discussion which fascinates me the most from a Less Wrong perspective is the use of words and categories. How do we arrive at a distinction between whether games are "variants" or "in the same family"?

Each category has strongly-bound traits and weakly-bound traits, such as "a matrix of regularly-spaced locations where game materials can be positioned". Even the category of abstract games has traits bound to it, like deterministic non-randomness.

Almost twenty years ago, I made several custom Shogi sets. The kanji characters meant nothing to me, and I had to look them up on a chart each time I took a turn. This was poor cognitive ergonomics, and imposed an unacceptable cognitive load, distracting from play. So I replaced them on my custom sets, with a pie-chart-like diagram on each tile, which showed where the tile can move. I omitted the kanji completely. This was the only way I could convince beginners to play it with me.

When I finally went to a Shogi event and showed the set to Shogi players, they felt almost actively-opposed to its existence. The reason they gave, was that they played Shogi to take part in a long tradition, and they wanted new players to learn the traditional kanji characters before they have an opportunity to even begin to learn how to play a strong game.

Keep in mind, in the two-thousand-aughts, I was a being of pure authenticity-- "authenticity" as defined as opposition to tradition, or ritual, or doing things because others are doing them. I was a full-on Microsoft-hating partisan of Linux. I called myself "transhumanist". I was president of the Logical Language Group, a standards body for a utilitarian constructed language designed around predicate logic systems. I kept talking about alternative calendar systems, and abolishing the penny and normalizing dollar coins. And more than anything else, I talked about atheism. A lot. A traditional approach to Shogi frustrated me because it was un-optimized.

So I hope you will understand if, since then, I've preferred the phrase "chesslike games", because "chess variants" treats FIDE chess as if it were a Platonic ideal. It's selected as the prototype of the category because it's familiar. Its mind-share makes it the equivalent to what Facebook is to social media platforms. Are chess variants "variants of chess" instead of "the chesslike family" because we have a traditionalist's loyalty to FIDE rules? It's not like a platonic solid. It's not.

And honestly, why should it be? Is Go a "Tic-tac-toe variant"? Tic-tac-toe is closer to a Platonic solid, from the criteria of simplicity. So are Symple, Hex, and the rest of the family. But we don't use minimalistic simplicity as the criteria. I think I see mind-share as the most common criteria. And this is why you get into pathologies like "true" 3D chess, which takes several days to play, it's impossible to catch the King, and it scales outside of the human mind. All in order to keep "true" to familiar FIDE orthodoxy, detached from the quality of the gameplay experience. A 3D chesslike game is a lot of fun, if you completely discard whatever you have to discard to make it playable and fun. "True" this, and "true" that, is a mental pathology.

I'm a published board game designer, and my design criteria only have to do with factors that make a game easy to teach, easy to set up, not take too long, and provide a depth that emerges from a rich set of simple interactions.

I'm interested in your concept of Chess feeling like it's complete, as if it has completed a platonic solid. You remedied its incompletions with the Cardinal and Marshal, but Omega Chess attempts to fill in the lack of leapers by introducing the Champion and Wizard. The design of Omega Chess was similarly motivated by this sense of completion, but it seems its design locates the gaps in different places from where the design of Grand Chess locates the gaps. Did you consider this? And how did you come to this conclusion?

I don't really feel like the Champion and Wizard are "missing pieces" the same way the Cardinal and Marshal are. There are a practically infinite number of pieces that could possibly exist, so I'd expect most possible pieces are not included in the game (Betza's Chess with Different Armies is a nice exploration of this). Omega chess doesn't even feel particularly complete- where are the pieces that can move exactly 1 square orthogonally or diagonally? If a champion is to rook as the wizard is to bishop, what is the knight to? I feel the correct completion of that is the knightrider, but there is no knightrider in the game. And we still have this weird gap where we have rook + bishop = queen, but rook + knight and bishop + knight are missing.

While I expect that most possible pieces will be missing from any variant, I happen to agree with Capablanca and Freeling that the lack of marshal and cardinal in orthodox feels weird. In orthodox, the queen is a strange wildcard that is unlike any other piece, whereas in grand chess the queen, cardinal, and marshal make a nice set, naturally extending the knights, bishops, and rooks.

Beyond any considerations of completeness, I also just feel more excited about the Cardinal and Marshal than most other variant pieces. They feel like fireworks to me, and I'd prefer a game with them over a game with a few new weaker pieces

Have you checked the game "the duke" out? After a piece moves, each piece has a different set of moves

Yeah, The Duke is pretty interesting. Along the same lines, look at Onitama.

Have you checked the game "the duke" out? After a piece moves, each piece has a different set of moves

You can significantly reduce draws in normal Chess by simplifying the rules and getting rid of 'check'. Just swap to the clean win condition of 'capture the opposing King'. In particular, this means all KIng+Pawn vs King endgames are won unless the friendly pawn and king are too far away. If KPvsK is winning a large number of other endgames are winning as well since they can be converted to KPvsK.  It has other implications. For example, all normal Queen vs Pawn endgames are won unless the pawn immediately promotes.

Aside from changing stalemates, I don't see how this changes anything. Checkmate is isomorphic to capturing the king, as long as no blunders are made

If the king needs protection, you can add the Pole from Pole Chess. The pole starts off the board, and can be dropped anywhere, or moved to any open square once in play. The pole cannot capture or be captured.

  1. This feels like it would be too much of a hack; while the ruleset I propose here is not maximally elegant, I do feel that it's a fairly natural solution to the problems that it solves, while this would enlarge the rules without being strictly neccesary

  2. I worry that the pole will get in the way of the king. Many checkmates I have seen, both in orthodox and crazyhouse, were due to the king's own pieces being unfortuitously placed on squares that the king would otherwise have been able to run away to. While most of the iconic smother-mates that I can think of were in orthodox since I play mostly orthodox, they are even more common in crazyhouse. I worry that the pole will smother the king just as often as it comes in handy to protect it

My guess is that this does indeed not strengthen the king enough to make drops not perpetually terrifying. It still has only fairly short-range moves, and you've added two more queenlike pieces to the potential attacking forces. I'm not more than say 60% confident of this, though.

It doesn't seem right to me to give the king a long-range move. To end the game, the king needs every single one of its liberties to be blocked off, and the shape the liberties makes affects how enjoyable it can be to try and pin down a king - the shape of the orthodox king's liberties is very pleasant to navigate (both as an attacker and defender), and I worry that buffing the king can end up with its liberties having a shape that isn't fun to play with. If the king has too many liberties, then pinning it down can be like trying to hold water; it's way too slippery.

The difficulty of checkmating a king increases exponentially as liberties increases - a naïve model: if any given square has probability p of being attacked by the opposing team or blocked, then there is a p^L probability of a king with L liberties being checkmated. Of course, this model isn't realistic, but the pattern holds, so even just a few extra moves can go a long way for buffing the king.

I'm not keen on giving the king long-range moves either. It's possible that if you don't debuff the other pieces, nothing much like crazyhouse will fail to be terrifying in something like the way crazyhouse is.

I think you're right that small increases in mobility can make a big difference to how hard to checkmate the king is. But the particular increase in mobility you've proposed doesn't change anything at all while the king is on the back rank, and doesn't change much while it's on the second rank. In normal chess, and I think also in crazyhouse, the kings typically stay on the back rank for their own protection, at least until the endgame. Maaaaybe with the extra mobility the kings won't need to stay hidden to the same extent, but I think being exposed is still going to be dangerous even for the enhanced king.

[Meta: I notice that in both of the threads where I've discussed things with you, all my comments have been downvoted. It looks as if the same has happened to other people who have posted comments disagreeing with things you've said. I find that, rationally or not, this makes me a bit reluctant to attempt to discuss anything with you. I suspect the effect on others may be the same. If you are indeed downvoting every comment you don't agree with, you may wish to consider whether the effects of doing so are the ones you want. For the avoidance of doubt, I'm neither complaining nor objecting; a policy of downvoting things one disagrees with is perfectly permissible, although perhaps a bit rude. Also for the avoidance of doubt, I am not claiming that you are downvoting every disagreeing comment; it looks that way but of course there are other possible explanations.]

I have downvoted a few, but certainly not all of the comments you made in response to me here and on my other recent post. On the other thread, I have downvoted comments that I both disagreed with, and felt were either too harsh or failed to understand what I was saying, without attempting to resolve potential miscommunications.

In the case of this particular comment, your comment was showing up at the top of the comments (due to your seniority on this site), but I felt that it wasn't the comment that presented the most useful information, so I weak-downvoted it to allow other comments to gain more visibility. I don't have any problem with the comment itself. Now that I say it, I'm not sure that I endorse that approach (not sure that I don't, either, but I'll reflect on this more tonight), but I definitely do apologize for any chilling effect that may have had, I feel bad that my approach has made you feel reluctant to engage in open conversation.

I will also note that there are several comments disagreeing with me on this post and the other one that I haven't downvoted - there are only 4 other comments (aside from you) that I have downvoted between these two posts, 2 of which were unhelpful and had questionable tones, and have received downvotes from people other than me, and currently stand at negative karma. The other two posts I weak-downvoted for the same reason as this one - they appeared at the top of the comments by default, due to the user's seniority giving their comments 2 karma by default, but which were not the most interesting comments.

This was originally posted on my shortform, and reposted as a top-level post on the recommendation of Raemon

There's some conversation about this post over on the comments on Hacker News, so if you want to hear more, check that out.

Bughouse Chess seems like an option. Bughouse is played on two boards. Pieces captured on one board can be placed on the other board. Bughouse is rather popular with strong chess players. Traditionally bughouse is played in teams of two. But nothing is stopping you from having one player play both of their boards. Given its popularity, it seems like an attractive option. Though you have to order the turns WhiteA, WhiteB, BlackB, BlackA. 

Changing the rules tends to neutralize acquired knowledge.  A strong club player is strong in part because he has an opening repertoire, a good knowledge of endgames, a positional sense in the middlegame, and recognizes tactical themes from experience.  Beginners tend to be weak players precisely because they lack those things, because they haven't  yet made the investment in time and effort to acquire them.

Changing the rules appeals to weaker players because it levels the playing field.

Of course, by saying this, I'm signaling that I'm a chess snob, that I have substantial acquired knowledge, and that I'm strong enough to play "real chess".

I mean, Capablanca was World Champion. Same thing with Bobby Fischer and Fischer960 chess