But the job of making O-O and O-O-O work in Fischer Random has been programmed in a very old fork of chess.js. There wasn't much interest at the time in chess960 so that fork isn't used much, and surely has bugs and issues outside of chess960 that the newer version of chess.js has fixed.

Like the author/maintainer of chess.js said, if there was sufficient interest he would merge the FRC code back into chess.js. I believe that's the best shot at making it work. I should sign into my GitHub account and revive the issue arguing that the two best blitz players in the world played an epic match, so we can no longer regard this variation as unimportant. Perhaps he'll make it a priority. If chess.js gets fixed it's a simple matter of replacing one file for another and instantly Olga would be processing chess960.

Another small issue is that if you examine the PGN from the recent tournament, the FEN is highly irregular, notably with the designation of kingside/queenside castling. For example, if the rooks started in their normal positions, the initial FEN would not end in "w KQkq -" but rather in "w HAha -". (Because you see, as Fischer himself pointed out, it's no longer appropriate to call it "kingside and queenside" since the king might start on the left and the queen might start on the right.) So the new terminology is "a-side castling" and "h-side castling". To make matters worse they change the letters depending on where the rooks start out (I noticed a "DAda" and a "GBgb" as well.) There's no reason we can't keep K and Q and just understand that for historical reasons we regard the K symbol to mean "white's right hand side" and so forth.

Computer technology is jam-packed with traditional conventions that lost modern day meaning, and there's nothing wrong with that as long as the software still functions. Cf. the LISP commands CAR and CDR.

I disagree, though probably not enough to hold up my end of an argument about it.

Games for replay:

http://www.frchess.com/live-games/

PGNs available from here:

>

Agreed it's the best solution, but we still face the problem of games with "legal" illegal moves.

<So the new terminology is "a-side castling" and "h-side castling".>

I don't see any additional functionally for the cost since Q-side = "a-side", K-side = "h-side". Maybe I'm missing something.

* * * * *

I was thinking we need to extend the PGN standard to formalize/accommodate things like Fischer960, and especially "legal" illegals moves (I see a practical need for that as a big priority to accurately reflect classical play).

I like doing the low level stuff a lot, and actually did some for SCID. Even bitboards for my FEN program. But I understand wanting to rely on an underlying framework/library.

Still, Olga could be a stand-alone piece of code - the kernel must not be that much work (cf my FEN program).

You're right, but there are certain cases where FEN falls over with chess960 (see the example here https://en.wikipedia.org/wiki/X-FEN...). X-FEN was developed to be backwards compatible with FEN, and therefore regular FEN is all that is needed to describe any chess960 game starting position (and the majority of in-game positions too). The special X-FEN cases are required when there is some ambiguity about which rook is the castling rook. If you're interested, you can read more about it here https://chess960.net/start-positions/

The reason for the redundant new-terminology is only because the queen's initial square can be either K-side or Q-side, and I guess people just wanted to be pedantic.

Ah, I see - this for FEN after the game has begun, say when one rook moves back to the 1st rank on the other side of the king. Then the FEN loses the info of which rook moved and which didn't.

However, the only legal castling would be with the nearest rook, as the unoccupied square rule still applies.

Am I missing something still?

And yes, 960 has the problem and classic doesn't.

You can use the normal FEN to forget previous moves in a recorded 960 game - since you can assume the remaining moves are all legal.

Remember though that FEN doesn't say whether castling immediately is a valid move or not, but only if castling rights exist. Those rights can still exist even when there are pieces in the way. For chess960, two rooks on the same side as the king can cause a problem if castling rights are still available on that side (how to know which rook is the castling rook, and which one moved over there during play?).

<zanzibar: You can use the normal FEN to forget previous moves in a recorded 960 game - since you can assume the remaining moves are all legal.

But you can't use a FEN to set up a 960 game to play in a forward sense, at least in the most general case. You would need an XFEN as <drleper> said.>

Players are forced to rely on their talent and creativity (instead of rote memorization of opening lines)

That's simply not true.>

If 960 becomes the standard, this "advantage" will quickly go away and players will memorize 960 theory. It's 960 times more? So what? Modern theory grew by more than that in the last century - think of it, in Nimzo'a times they could cause controversy by playing a novelty on move 3.

Proof needed.>

But regular FEN can be used to truncate any 960 game's movelist beginning, as I said last night. The assumption is that all the subsequent moves are legal. It's not a big issue, and doesn't have much functionality, so no need to belabor it.

For people wondering about the difference between S-FEN/X-FEN/FEN castling rights - I think this page is useful:

https://www.chess.com/clubs/forum/v...