Chess18 -- better than Chess960?

Introducing Chess18:

Chess18 is a hybrid of traditional chess and chess960, an attempt to better balance the virtues and weaknesses of each.  Chess18 uses only the 18 starting positions from chess960 that have, like traditional chess, the rooks in the corners and the kings on the e file.

For at least 100 years – Lasker was the first we know – elite players have complained that standard chess was “played out.”  Capablanca became famously bored of standard chess and spent much effort developing variants, including a proposal to swap the starting position of the knights and bishops “to inject new life into the game.” Fischer’s rationale for creating chess960 was centered on his complaint that chess was played out and rewarded brute memorization at the expense of creativity and imagination.  Kramnik has been experimenting with no-castle chess and other, more radical variants. And when Carlsen disappointed the chess world announcing he would not defend his world champion title, most of his rationale stemmed from the process – months of memorizing computer analysis – being joyless.  Obviously, many players feel chess is tired.

Chess960 has been generally well received and is currently the most accepted variant. But critics have identified drawbacks that limit its widespread adoption. The primary complaint is that eliminating opening preparation altogether kills an important and treasured dimension of the game. Another criticism is that many starting positions have gimmicky castling patterns, and positions that feel too far removed from traditional chess.  Commenting on Fischer’s invention, Kasparov proposed choosing perhaps 10-20 starting positions from among the 960, allowing for new positions but few enough to be played with sufficient frequency to each develop their own opening theory.  Many other players and commentators have made similar criticisms and suggestions.

Chess18 attempts to synthesize these ideas by accepting Fischer’s framework, modified by Kasparov’s suggestion for fewer positions and, by choosing the positions most similar to traditional chess due to identical castling moves, conveniently ending up with a number of starting positions within Kasparov’s suggested range.  The rules of play are identical to chess, and the legal starting positions are easy to remember and explain, a crucial consideration for the world’s most popular game.

Chess18 actually expands rather than eliminates opening theory, but expands it so dramatically that it ensures even elite players will be out of their preparation within a handful of moves. (If players were to spread their opening preparation across 18 starting positions, their knowledge of any position would be only 1/18th – less than 6% – of the theory they currently know of standard chess.)

While I believe in the promise of Chess18, until now the potential has been wholly speculative, and I’m excited to finally submit the format for play by titled players at Timber Moose Chess18.

Chess18 attempts to synthesize these ideas by accepting Fischer’s framework, modified by Kasparov’s suggestion for fewer positions, and, by choosing the positions most similar to traditional chess with identical castling moves, conveniently ending up with the number of starting positions within Kasparov’s suggested range. The rules of play are unchanged, and the legal starting positions are easy to remember and explain, a crucial consideration for the world’s most popular game.

Chess18 actually expands rather than eliminates opening theory, but expands it so dramatically that it ensures even elite players will be out of their preparation within a handful of moves. (If players were to spread their opening preparation across 18 starting positions, their knowledge of any position would be only 1/18th – less than 6% – of the theory they currently know of standard chess.)

While I believe in the promise of Chess18, until now the potential has been wholly speculative, and I’m excited to finally submit the format for play by titled players at Timber Moose Chess18.

What about Chess 18 and the inclusion of mirror image #534, i.e. that the king starts on the queen's square. That is, approximately double of 18 -> 36, If I'm not much mistaken.

Good question.  I initially considered using the positions with the kings on d, and adopting the castling rule that the king moves 2 squares instead of the 960 castling rule that the king moves to either c or g no matter the starting position (since one of my goals was eliminating the gimmicky castling patterns in 960).  My thought was that even though this made the positions with king on d identical mirrors of the position with king on e, it would nevertheless create a cognitive challenge that would also undermine excessive opening preparation.  But during discussions with GM Larry Kaufman weighing the tradeoffs I decided that having simple rules and staying truer to traditional chess is more important to win widespread adoption (my primary goal is to have more chess events that play non-Chess1 starting positions) than is the benefit of modestly increasing the cognitive load of mirrored positions.  I believe 18 positions is sufficient to spur creativity, and think Chess4 -- the four positions with king on e and queen on d -- would be wonderful, too.  I would love a future where the distribution of chess tournament formats were roughly evenly split between Chess1, Chess4, Chess18, and Chess960.    

The castling outcome is not mirrored. I lost a game #534 because of that.

I know about it. The point is, as long as you know about it, it's not a problem

To make Chess18 work for match play at classical time controls, since the length makes it infeasible to ensure fairness by playing the same starting position twice, alternating white and black, I think each player should choose, say, 12 starting positions that they will play as black.  For an event like the world championship I'd have the players announce their 12 positions far in advance.  The position for each game could still be randomly chosen immediately before each game, and a match of 12 or 14 games could stipulate that no position would be used twice and there would still be 6 or 5 possible starting positions going into the final game.         

I have no problem with the 960 positions that make castling really hard, but there is no question that some of the positions feel a little stiff, like with both bishops in corners, or both bishops on the far left or right.  Still, there were some games that got really interesting starting from those in the FR championship.

uh that does LITERALLY nothing as you quite literally just mirrored the board in its entirety, meaning that yes, technically it's 36 possible combinations, but in reality, you just mirror the board and "oh hey I've seen this position before"

“I have no problem with the 960 positions that make castling really hard, but there is no question that some of the positions feel a little stiff, like with both bishops in corners, or both bishops on the far left or right. Still, there were some games that got really interesting starting from those in the FR championship.”

I agree with this, though my complaint about 960 castling isn’t that it’s hard, but that it’s frequently gimmicky (e.g., moving the king up to *six* squares?!)

I consider Chess18 less an improvement on 960, though I do like it better, than as a compromise that can appeal to those who still prefer Chess1.

Players being in their preparation 15 moves in is so dull — it’s like watching a pi memorization contest.

@jo857294 Now you forgot that castling is not mirrored. 

She is Al ok

Nice

Ok

Chess18 is actually chess24 (1•2•3•4)

I agree that moving the king six squares can seem excessive, but I think there are good reasons to justify this.

1. En passant is kind of the same way. It's an ad hoc rule to save the mechanics of chess while allowing pawns to move up two. It is the one capture in chess where a piece doesn't land on the square of the captured piece. Yet, no one complains about this. I think a similar justification can be provided for unconventional castling in Chess960.

2. Having castling always end up in the same position as in standard chess ensures that Chess960 is as chess-like as possible which makes it a good successor to the game.

3. This sort of castling makes the game more exciting than if castling was always the same. It expands the geometric possibilities which I think is a good thing.

I've also thought about the possibility of Chess204. These are the subset of Chess960 positions where the king is fixed on the e-file. This means that for castling to occur, the king will always move 2 squares in either direction, regardless of where the rooks are placed.

However, I still prefer Chess960 because I don't like the idea of fixing pieces. I think more diverse castling options (and positions) makes the game better, and 960 positions is probably preferable to 204.

My problem with Chess18 is that there doesn't seem to be a principled reason to fix the king and rooks (at most the king would need to be fixed), there is much less positional diversity than in Chess960 (which makes it less interesting to me), and Chess18 allows for some opening theory which would probably be problematic. The goal of Chess960 was to eliminate (or at least greatly reduce) opening theory. Chess18 seems to exacerbate this problem, not fix it.

I've been thinking of ways to improve chess or Chess960. I've been thinking about it for a while, but I think that Chess960 is the best. Fischer was a genius. It has the simplest rule-set for randomizing the pieces, all of the positions are relatively equal, and the expanded geometric possibilities are amazing.

Yesterday i found out about Kraton Chess. The King is fixed on the e-file, otherwise the remaining pieces are shuffled and castling is legal with rooks on other squares if i'm not mistaken. With fixed K, the positions increase to 96 a considerably increase compared to 18.

Asymmetric Chess18² give 324 but some express concern that people are not likely to accept uneven opening setups with various evaluation which give small or large edge.

FRC is not a perfect solution as it has some problematic sides compared to the classic start position which most people are accustomed to.

This Chess 18 is an ingenious proposal, but there's really no need to fiddle around with the starting position in this way. Standard chess still has plenty of life in it, although it is true that at the elite level engine prep has become a problem.

To refresh standard chess, my proposal is to have opening ballots. For example, there are over 1,000 6-ply openings which could be used, allowing for great diversity in the opening. Obvious early losses would be weeded out by computer analysis. The ballot for a game would be selected randomly when the players sat down at the table.

A ballot system doesn't do away with standard chess, but puts more emphasis on over-the-board skills, with versatility being a key quality. Players could not just trot out an opening line a computer found for them.

I honestly don't know why this ballot route has not been tried instead of all this playing around with variants. Incidentally, ballotted openings have worked very well for more than a century in British/American Checkers.

Another persistent issue with the ordinary position is that the same positions often occurs time after time. It depends on how you look at it, but it gets a bit monotonous in the long run if you ask me. Isn't it better to have a completely new position on the board, where creativity and talent are used rather than repeating the same position time after time.

My problem with chess 960 is that the pieces must mirror each other. I'd say, regardless of what 1st rank setups are legal, that the sides shouldn't have to mirror each other. I don't buy that it makes some positions automatically winning/losing because of possible imbalances, not with all the pieces on the 1st rank still blocked in with their pawns.

I take your side with your point there. I believe that asymmetric 960 does not necessarily mean that people can exploit it there and then. A machine and a human can assess it differently.