Sicilian weakness

@1

The best response to a flank attack is a central counterstrike, like ...d5 or ...Ne5.
You do not need to succeed in the counterattack, if you can just defend your king you often get a good endgame because of your central pawn majority.
Look at some games by Kasparov
https://www.chessgames.com/perl/chessgame?gid=1333302 

don't use a book, use the lichess database! first off, it'll give you theory for EVERYTHING everyone's playing in the real sub-2000 world AND you can see how lines PERFORM! grandmaster and ESPECIALLY stockfished approved lines often are the WORST performing over the board while "unsound lines", sometimes up to 5 points bad kick the most butt because even 2000s don't catch the trouble they're getting themselves into.

the great thing about using the database instead of book theory is you can track down most of your opponents' mistakes as they reveal themselves in the stats. there are "unsound lines", like the "zeller ice queen" variation in the blackmar diemer gambit that have something like 70% win rates as soon as move 4! get trappy! 

haters will keep vomiting grandmaster ideas about gambits that have absolutely ZERO CREDABILITY in the real world! popular gambits often have win rates of +8%!!! "sound GM lines" tend to perform 48:48, or sometimes even worse because everyone's figured those out, but street fighter trappy lines? they catch even grandmasters off guard!

check videos by gotham chess and john schrantz out on youtube. they are a GOLD MINE for crazy gambits that kick much butt in the real world. I'd like to see any haters here face EITHER of those 2 over the board with their tired old karpov lines. pawns don't matter when you have the initiative.

i'm going to take a look at your games to find what your doing in the sicilian to give you an example of something that scores better if i can.

OK, you play the sheveningen. and this is your most recent game

2...d6 scores 48:47 in favor of white and by the time you get to 5...a6 in your theory, white has an edge in EVERY LINE. small advantage get bigger. 2...e6 scores 46:49. that's already 3% better. following the (most popular) "main line", the 4 knights variation is STARTING to kick butt in most lines by move 5!

an early Qb6 (forgot which move) performs even stronger, especially in the main line, but the  2nd most popular reply by white favors white several percent if you want to win 61% of your games in the most common line, and overall, then the Qb6 line might be for you, but I followed the stats to the 4 knights where it's doing good against everything

THIS LINE has a 6% advantage overall in ALL games (the 772,318 sigma line at the bottom) it isn't until you get to the 4th most popular line that white starts performing better, but that'd be about 1 in 5 of your games after the top 3, and black continues to do well the further down you go. 

a nice thing about rolling your own books (or finding awesome youtube videos that have ALREADY found killer lines GMs & haters will try to steer you away from) is that you can chose your own priorities, do you want to open the position or close it? how much risk are you willing to take? how deep do you want to go in your book? would you rather play a line that a couple % worse performing that has only 1 real reply, or would you rather learn several lines that perform stronger? geometric progression is a mofo there! I'll take the forcing line that goes 5 moves before branching starts myself.

when you follow the strongest performing lines, eventually you'll find that the most popular replies perform the worst, down to even mates in 1 by move 8! (if you play double edged gambits anyways). 

the strongest performing 5.Nbd5 above is a "secret weapon" line for white. the further down you go for stronger performing lines, the more of a surprise they'll be to your opponents, but the down side is that you'll have fewer games from which to draw theory. 

lines that are the most popular AND perform better for black are "tried and true". the herd knows what it's doing most of the time, until a new secret weapon climbs up as the next main line.

it's tough trying to find the strongest lines in the sicilian because it's soooo bookish most lines perform roughly equal until you really dig down, but i found a line that's stronger performing without much effort here. it's not necessarily even what you'll want to play, but small advantages lead to bigger ones as you find all of your opponents' most popular blunders.

sometimes too, you should follow lines that are close in stats a few moves to see what happens down the road. then, you can find those hidden secret weapon moves that all of a sudden are 10% stronger than another line that started out 2% stronger, but stays that low (48:48 style) the same number of moves.

chose lines that make sense TO YOU. if you keep following the trail of crying 1.e4 players, the stats improve. you can look ahead to the next ply and even the next ply after that for tie breakers. 

if you REALLY want to kick butt though, I'd suggest looking into creating a gambit repertoire. as a king's gambit player, "reverse king's gambit" lines like the calabrese counter gambit and others that play an early ...f5 can lead to huge advantages in the main lines, but you'll need to cobble your theory from different variations. it's a lot of research, but who wouldn't want a repertoire that MOSTLY performs better, except for a few troublesome lines (or EVERY LINE following your current book) that include POPULAR line stomping secret weapons that score as much as 80% wins?

gambits might not be for you, but if you're an attacker, like me, they are soooo delicious!

hope that helps.

oh... you can create books and train against them at chess tempo too. it's a lot of work creating a complete repertoire, but you could always hit the ground running if you start out just learning the first 5 to 8 moves in your lines. after that, your book gets huge fast.

risk taking gambiteers like schrantz, gothem et. al. see a line's PERFORMANCE over the board as way more important than a stockfish or grandmaster evaluation, even playing grandmasters or stockfish! they can fall victim to traps too, especially as they never see what we see on the streets. it's nice that you can always use SF as a tie breaker. sometimes, it will find moves that are 5 points or more better everyone overlooks, but it also finds a lot of weird computer moves players do terrible at as they don't think like computers. you'll see a lot of stockfish choices losing badly if you dig through stats long enough.

hope that helps.

P.S. if you stick with the sicilian, you better study the smith morra gambit, because that's how I'LL come at ya. 1. e4 c5 2. d4 cxd4 3. c3 dxc3 4. Nxc3 is 52:44 in 1.5 million games. not MY games, but overall. there's that 8% gambit edge haters like to pretend doesn't exist, but the numbers don't lie.

indeed. The only numbers I do believe don't lie is 1382 > 1234.

It is really depended on which variation of the Sicilian you choose to play. If you are a Dragon Sicilian player, you have to be very careful. As white will usually just roll black down with a big kingside attack. But if you play something like the Taimanov, you can get a solid position which is still complicated. So it is really depending on which variation you decide to play. 

Bur if you still want to play a complicated opposite side castling game, you have to know that not every counterattack works. As white has the extra tempo for being white. So if you see that your counterattack is not fast enough, try to create play in the center.

get better

As Black in the Sicilian I typically just leave my King in the center.

That's an overly generic statement.

A lot depends on which line of the Sicilian.

Dragon, Accelerated Dragon, Closed, etc, Black almost always castles Kingside.

Najdorf, Sveshnikov, etc, Black almost never castles.

To add a few more words to what blueemu said, very generally the idea is

prepare to castle but dont actually castle until you get counterplay in queenside or center started.

Sicilian is very concrete so general ideas are less important than  situational specifics.