Why Do(n't) Chess Variants Work?
An enquiry into the properties of chess variants

Why Do(n't) Chess Variants Work?


What's the most fun you've had with a chess variant? Perhaps like me, it's playing bughouse, which is not only innately fascinating, but a good way to make friends and interact with people who are not solely interested in taking all of your elo points.

Here, I'm interested in two questions. Firstly, what, if any, are the shared properties which make games like bughouse, crazyhouse, king of the hill, 3 check, such interesting games, and as successful as they are? Secondly, why aren't these games played more--played by strong grandmasters, played for monetary rewards, and played at longer time controls? 

1. Continuity with Chess

When we are discussing chess variants, it is one thing to discuss the properties of a game, and another to discuss it as a chess variant. In the first instance one might discuss things like balance (does the first player get an unreasonable advantage?), depth (is the game fun enough, or is it too easy), duration (is the game of the right length?) and other properties, which I discussed in an older blog post.


These are criteria that might apply to other games of complete information. But what I'm asking is why variants of chess work in relation to chess. 

Sibling variants

Strong variants are different from chess, but related. Take for example, chess 960. Everything is the same as in chess, except for the starting position. You fight for control of the center, and then the positional play and tactical traps are more or less the same--forks, skewers, pins--that kind of thing. Chess 960 is like a sibling of chess. Strength in chess is closely tied to strength in chess 960, but as the recent match between world champion Magnus Carlsen and world top ten Wesley So shows, one's ranking in chess does not fix one's ranking in chess 960. Chess 960 is close enough to chess for chess players to have a stake in being good at it. One might reason that since the only difference lies in the initial setup, one should be flexible and versatile enough as a chess player to be able to deal anything 960 can bring.

An example starting position in chess 960

King of the hill retains the original starting position of chess, but adds an additional goal. One can win in all of the usual ways in chess, but added to this, getting your King to any of the center squares e4, d4, e5, or d5, is a win.

White is being crushed in chess, but here b4 forces a win. 

It's an interesting game, again very close to chess. As I understand it, more or less any opening which is viable in chess is viable in this variant. My impression is that strength in this variant is closely correlated with chess strength. In my games with titled players, I lost many games as though they were chess. Against these players, the uniqueness of KOTH did not even arise. Again, I would categorize KOTH as a sibling variant. It does not currently appear to matter very much to chess players, but this could in principle change. 

In the case of 3 check, it is a bit strange. After 1.e4 e6 it would be bad for White to play the natural move 2.d4, since Bb4+ is a free check for Black. Because the race for checks dominates the game and changes the opening theory and character of the game so radically, I wouldn't call 3 check a sibling variant. But on the other hand, since checkmate remains a possible and plausible way to win, and because there are so many other continuities, I would consider 3 check a step sibling of chess. The main problem with 3 check appears to be that the Black player needs to know a lot of theory just to stay in the game against a strong White player. 

Cousin Variants

In crazyhouse and bughouse, the drop rules take away much of the maneuvering you find in chess. But they are still close. Bughouse is perhaps closer to chess than crazyhouse is, since "chessing" is a very useful tool to have in a bughouse players skill set. If you are up time, you can ask your partner to wait, and play for a while as if you were in a game of standard chess. There are many situations where a bughouse game will reduce to a chess game on one board. 

Alien Variants

A very difficult to categorize variant is atomic chess. Captures cause the pieces to explode.

It's chess--everything moves in the same way--but it's very unusual. It has its own opening theory, which is bizarre from a standard chess playing perspective. Standard tactics go out of the window, and so does standard positional play. Virtually nothing Nimzowitsch wrote is applicable in this game. And yet it works, arguably. The only problem with it appears to be that White enjoys too large of an advantage.

Another variant you might consider to be in the alien category is Relay Chess. In this game, any piece (not pawns) which is defended by another piece, gains the abilities of that piece. 

Descendant Variants

A well known, but underplayed variant is Seirawan Chess. It is played on a standard board, with standard pieces, but adds two new pieces, the Hawk (Knight plus Bishop) and the Elephant (Knight plus Rook). (It is inspired by Capablanca Chess, discussed below.) These pieces may be introduced when a piece (not pawns) has moved off of its home square, onto that vacant square. The upshot of all of this is that it's standard chess with a lot of new life injected into it. For instance, you can play the Scandinavian defense and introduce the elephant to d8 as you play 2...Qxd5. 

And another example is that you can play a Sicilian as Black and introduce the Hawk to b8, which gives you extra coverage of the dark squares. The Elephant (Rook+Knight) may be placed on c8, or g8, where it can support g5. Moves which would be considered to create irreparable positional weaknesses in chess become viable in Seirawan, since you have more pieces to protect weaknesses with. 

I consider this a real strength of Seirawan Chess. It is familiar, and yet new. Much opening theory from chess remains, but it is interspersed with surprises. 

As another example of a descendant variation, consider the variant, Shako, invented by Jean-Louis Cazoux. It is played on a 10x10 board and introduces two new pieces, the elephant and the cannon. Both of these are taken from Chinese Chess. (The elephant is adapted slightly). 

It retains the initial starting position of chess, and builds around it. The new pieces are minors, similar in strength to Knights and Bishops, and so the game is very chess-like. 

I take it that bughouse and crazyhouse are the most popular of all of the variants (but I have no actual numbers to back this up). If this is true, what accounts for it? Broadly speaking, dropping pieces is intuitive and fun. The drop rules are basically simple, but completely game-changing. Parsimony is important. Too many additions to chess, or too many convoluted rules, will cause a variant to fail. Any exception to a rule must be there to solve a problem. Take for instance, the en passant rule. It is rather ad hoc, but it keeps the board from becoming too closed and drawish. Large variants tend to be too excessive in the eyes of the vast majority of chess players. 

Just as you can organize a set of books alphabetically by title, author, or by the year published, we can organize chess variants through different perspectives. 

2. Novelty

Being close to chess is a virtue for a chess variant, but it is not the only criterion. The best variants take a simple but compelling idea. What if we change the starting position? What if we keep the starting position, but change the winning conditions? What if your captured pieces become your own? What if you can introduce two new pieces into the game? Anything very complicated is probably going to make an unplayable variant--since chess is already complicated to begin with. 

Some of the most interesting aspects of variants were probably not planned in advance. Take, for instance, the role of time in bughouse. The role time plays in bughouse is a byproduct of a number of stipulations. (The two boards are not required to move in step.) Another interesting feature of bughouse is that it is perfectly balanced, since there are two boards, and each team has one player playing Black, and another playing White. This is needed in order for the game to work, but was probably not a conscious part of the design. 

In four player chess, the most interesting thing is the need to take into account more than just moves. A strong four player is sufficiently active to get ahead, but also conservative and solid enough to avoid being ganged up on by the other three players. 

Of course, this makes it very hard to see in advance what a good variant would look like, since it is virtually impossible to design a variant with the features which make it so interesting. 

3. Problem Solving

Ultimately, variants should solve a problem. Chess 960, which is perhaps financially the most successful variant, solves the problem of creeping opening theory. And by extension, it solves the problem of the draw death of chess. In every game, players are faced with new problems. Strong (or simply booked up) players executing twenty or so home-cooked moves is not a problem here. But Chess 960 is not perfect. Some of the starting positions are too awkward and unharmonious, and though the awkwardness is mutual, it can sometimes favor the first player. Arguably, beauty is a key ingredient to chess, and the clunky inelegance of some of the starting positions is a demerit. But this is to my mind, a good thing. As old problems are solved, new ones arise. 

The third World Champion, Jose Capablanca, proposed a variant which increases the size of the board (10x8) and adds the Archbishop (Knight+Bishop) and Chancellor (Knight+Rook). As a game, it's playable enough. But it feels artificial. The original position is not the result of hundreds of years of play, but of the play-testing of Capablanca himself and a few other strong players. A chess player is not as at home here as in Seirawan Chess, or many of the other variants. Why is it 10x8 anyway? 

If I had to categorize Capablanca Chess in the style of the taxonomy above, I would call it a Rival Variant (which seems contradictory), since it was designed to usurp the place of chess rather than as a game which can exist alongside it. Capablanca reportedly wanted to stave off the draw death of chess, and to stave off the problem of creeping opening theory. But since Capablanca proposed his variant, our collective understanding of chess deepened as players continued to study it, and to play fighting games, adding to the canon of games a variety of personalities and playing styles, none of which Capablanca could have predicted, since it was his contention that within a few years, even amateurs would figure out how to force a draw against stronger players. This is a laughable proposition now. Capablanca's mistake was in trying to predict the state of future knowledge, using current knowledge. Capablanca seems to have been premature, as chess went on and flourished. 

4. Success and Failure of Variants

I wonder if Capablanca's efforts were self-undermining. Since chess has been declared to be stagnant since the 1920s, in plain contradiction with the facts, perhaps this made the chess community wearisome of such claims. But this is just speculation.

In fact, variants are more successful now than they have ever been, since there are communities organized around them, and importantly, databases of games. A canon is important--it makes the difference between an abstraction and a real game. 

It seems that one of the main no-go areas for both chess.com and lichess is with the introduction of any nonstandard piece. I have no idea why this is. Is it a technical reason, to do with coding, or a social reason--would the new pieces frighten too many players? I have no idea. 

Just as a side note, where does Kramnik's suggestion that chess simply be played without castling fall into all of this? It looks like a strange step backwards. Why not instead, for instance, allow flexible castling? 

 In this situation, White can move his King to any square on the back rank, and place the rook on the other side of it, adjacent to it. For example...

Or perhaps...

It seems to me that giving players more options, rather than fewer, would be a better way forward. 

However, FM catask and IM opperwezen have explored chess without castling, and you can view the games, with catask's annotations, here: https://www.chess.com/blog/catask/no-castle-chess 


I am curious to hear from any readers. Which variants do you enjoy the most? What variants would you like to try? Is it the high rate of draws, the over-analysed openings, or something else, that makes chess stagnant? Or is chess just fine the way it is?