Chess18 -- better than Chess960?

I don't agree. I think asymmetric starting positions have too much potential for unfair advantages for one side. One thing I like about chess is that both sides start with the same exact position.

Then why do people favor it over normal chess?

I favor random starting positions over normal chess but only if they're symmetrical (which they typically are).

So why is that better than normal chess if they are just as symmetrical?

New positions generates more possibilities which I like. Symmetry makes sure those possibilities are fair for both players.

Yeah but then it's too easy to get symmetrical positions in the opening, I prefer more random piece positioning.

Random is good but not if it leads to unequal starting positions which asymmetrical positions will sometimes.

A few more starting positions might be better, and cherry picking positions with equal eval might be better.

not equal, +0.2

we need to keep the inherent first move advantage in any position list

From the Timber Moose C18 event website.

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.

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.

I am also interested in Chess18 and come up with a formula for converting any number from 0 to 17 into a unique starting position.

Chess 18 starting positions

Determined by converting the selected number (0 - 17) to base3 trinary, then using those digits to place the Bishops and Queen on different squares according to simple rules.

1. Setup the Rooks and King on their standard positions.
2. The first digit of the base3 number (9's) sets the Dark Square Bishop position by counting left to right available dark squares, a 0 here will be the 2nd available dark square in g file, a 1 puts the dark bishop on c file
3. The middle digit sets the Light Square Bishop position, 1 = b file, 2 = d, 0 = f
4. The ones digit puts the queen in one of the remaining 3 squares, counting left to right
with 0 being the last open square.
5. Place the 2 Knights in the 2 remaining squares.
*. The above rules apply for Whites pieces, Black pieces get set on the same files as White.

dec : base3 : R???K??R 
 00 : 000 : RNNQKBBR 
 01 : 001 : RQNNKBBR 
 02 : 002 : RNQNKBBR 
 03 : 010 : RBNNKQBR 
 04 : 011 : RBQNKNBR 
 05 : 012 : RBNQKNBR 
 06 : 020 : RNNBKQBR 
 07 : 021 : RQNBKNBR 
 08 : 022 : RNQBKNBR 
 09 : 100 : RNBNKBQR 
 10 : 101 : RQBNKBNR 
 11 : 102 : RNBQKBNR 
 12 : 110 : RBBNKNQR 
 13 : 111 : RBBQKNNR 
 14 : 112 : RBBNKQNR 
 15 : 120 : RNBBKNQR 
 16 : 121 : RQBBKNNR 
 17 : 122 : RNBBKQNR

For Chess36 mirrored, use numbers above 17 to flag as mirroring and subtract 18, then set the King to d file and follow the same rules as above but count right to left, and the base3 digit used for light and dark bishop positions switches because the King is on a light square.

Mirror - Chess36
 18 : RBBKQNNR
 19 : RBBKNNQR
 20 : RBBKNQNR
 21 : RBQKNNBR
 22 : RBNKNQBR
 23 : RBNKQNBR
 24 : RBQKBNNR
 25 : RBNKBNQR
 26 : RBNKBQNR
 27 : RQBKNBNR
 28 : RNBKNBQR
 29 : RNBKQBNR
 30 : RQNKNBBR
 31 : RNNKQBBR
 32 : RNQKNBBR
 33 : RQNKBBNR
 34 : RNNKBBQR
 35 : RNQKBBNR

I don't like the idea of having any fixed pieces. 18 positions still leaves room for problematic theory. Plus, it greatly reduces the geometric possibilities of 960.

The whole idea with Chess 960 or 18 is going in the wrong direction. Why do we actually want a different version of chess: to avoid openings theory. But in my opinion, a new version of chess should also go a bit further and make the game better. At the very least, castling, which is a nonsensical rule, should be eliminated: as a result, the rule is complicated and is probably responsible for a higher number of draws. But the castling rule is particularly strange in Chess 960, where, for example, the king with b1 and the rook on c1 jump to the squares f1,g1 ?

The second rule to abolish concerns the stalemate. In all similar board games (including the original and other versions of chess), stalemate is a game loser.

So the castling should be cancelled, the stalemate cancelled, and the eventual chess960 should be converted to chess1440 (without castling).

From talk to practical reality is something else. Everything must be analyzed and remembered. Most people have forever enough to stick to the usual position. I have my doubts that 960 players has the resources to start building up more theory on positions that occur once every jubilee year.

How about additional 18 possible positions with the King and Queen changing places. Shouldn't that be enough ?

17 positions (excluding the regular one) with additional 17/18 with the #534 960 aka mirror image starting position.

I think the solution to the whole king and rook thing is to just eliminate casting in 960 positions where they aren't on the correct squares. And abolish the symmetry requirement. I highly doubt any position with all the pieces on the 1st rank could ever leave one side with an advantage.

I spent some time running evals, here is my data.

dec : base3 : positions: Stockfish 14+ NNUE Eval (lichess)
00 : 000 : rnnqkbbr : +0.3 Depth 35
01 : 001 : rqnnkbbr : +0.3 Depth 39
02 : 002 : rnqnkbbr : +0.5 Depth 33
03 : 010 : rbnnkqbr : +0.3 Depth 36
04 : 011 : rbqnknbr : +0.5 Depth 33
05 : 012 : rbnqknbr : +0.2 Depth 41 Cloud
06 : 020 : rnnbkqbr : +0.5 Depth 38
07 : 021 : rqnbknbr : +0.2 Depth 35
08 : 022 : rnqbknbr : +0.3 Depth 36
09 : 100 : rnbnkbqr : +0.4 Depth 35
10 : 101 : rqbnkbnr : +0.3 Depth 33
11 : 102 : rnbqkbnr : +0.3 Depth 53 Cloud
12 : 110 : rbbnknqr : +0.2 Depth 33
13 : 111 : rbbqknnr : +0.2 Depth 35 Cloud
14 : 112 : rbbnkqnr : +0.4 Depth 40 Cloud
15 : 120 : rnbbknqr : +0.4 Depth 33
16 : 121 : rqbbknnr : +0.4 Depth 43
17 : 122 : rnbbkqnr : +0.5 Depth 33 Cloud

I have now implemented Chess18 as a Zillions program: Chess18 - Fischer Random with only 18 positions. (I have tweaked Zillions so that it plays decent chess.)