Hi everyone,

I've recently discovered a disturbing leak in my game - I'm lost when there's a close choice between castling kingside, queenside or even (in close positions) not castling at all (putting the king on e7 or d7 to connect the rooks, which has benefits too because, when the centre is solid and the play on both flanks of the board, I can put both rooks faster to the flank where they're needed, depending on the situation). Wrong castling decisions must have lost me many points already

I'd like to discuss Saemisch KID specifically, putting the wall of text in a spoiler to reduce the digression for those who don't rely on it with White:
An example game where O-O-O in the h3 KID went wrong for me has been discussed here.

A troublesome opening with Black as to castling seems to be the Black side of the Grand Prix Attack because it's well-known that O-O runs into Qe1-g3(h4) and a pawn storm. So I reckon that, as long as I put a pawn chain on the light squares (d5, e6, f7, g6, h7), whose primary purpose is to blunt White's b3 bishop, it has a pleasant side effect of solidification of the centre, so as long as I keep the c-file under control, I guess Kd7 or Ke7 is slightly better than O-O even despite the missing Black c-pawn. This somewhat resembles the Caro-Kann, where Ke7 is played quite often because the centre is closed.

Finally, a general issue is whether to castle or not after an early transitioning into the endgame. On the one hand, rooks are connected faster this way, but on the other hand, the king needs to be centralised and ends up a bit offside on g1/g8.

I guess, whether to castle or not in a complex endgame really depends on the coordination of the other pieces, i.e. whether the central king march disrupts it, and also on the opponent's coordination, i.e. whether the king is at risk of a non-mating attack which would however grab some material by means of double or combined attacks.

Let me share an example position for this case. (It's White to make the 17th move.)



I'll post at least a couple more example games when they finish

Please share some insight on the factors that you find important in the decision on where and whether to castle. Thanks!

The games in question didn't wait to finish , so let me post the critical positions where I was contemplating the castling direction (or abstinence from castling).

I looked over the diagram games, and you lost because of concrete oversights -- not incorrect castling decisions. Can you link a couple of games where your castling decision was objectively bad (and not "bad by titled player standards" either )?

You're playing correspondence games, so at the very least you can calculate a lot of variations and get a feel for what the position calls for.

In the first position 0-0-0 looks like the right plan, followed by a K-side advance. Black's counterplay with c5 and b5 seems to just drop the d-pawn for insufficient compensation.

In the second position I'm not sure what's going on. Whatever side you castle on it looks like you are going to run into some problems when White prepares for c3. I think the best move might be 1...f6.

Yep agree with 0-0-0 in first position since black already has weaknesses on his K-side and if you castle 0-0 they won't be exploited.

2nd position already went wrong for you imo unless a quick f6 now works(I think either don't play d5 with g6, e6, or just play d5 but don't fianchetto and creating f6 weakness), and castling either side won't change your ev much.

While f6 sounds like a decent plan too, Stockfish recommends to advance the a-pawn, bring the bishop to c5 and the knight to c6 (or maybe to f5, depending on what White plays). The thing is that if Black doesn't play a5 now, White can play a4 and Black won't be able to respond by b4, hence the a4-e8 diagonal or the a-file will open up to White's advantage.

The biggest weakness on Black's kingside is not f6 but f7. Typically in the Grand Prix, White sacs a pawn on f5 in order to open up either the bishop's diagonal, or the f-file, or the g-file, depending on the way Black captures the pawn. That's why castling is worse than it seems at first glance.

Regarding the prehistory of the position, g6 and Bg7 was played long before, while the purpose of d5 was to shut White's bishop out, because of the abovementioned f5 sac threat. White normally takes en passant (exd6), but this opponent chose to keep the centre closed and double my pawns.

This might be the critical position of the opening (after White's move 8):



In Pujari vs Prathamesh, Black chose to castle first, allowing the doubled pawns. After 9. Nxd4 cxd4, White put the knight on e4, which allowed Black to play d5 with a tempo and then, after White played 10. Ng5? instead of exd6, elegantly undouble the pawns by playing Qa5 (forcing White to move the c-pawn to save the b5 bishop).

However, there were 5 other games with that position shown by Chesstempo, and the 8... a6 line, where there were 3 games, all won by Black, instead of 1, sounded more legitimate. Though I saw that both games with 9. Bc4 d5 10. Bb3 Nec6 had been won by White, I thought that was because Nec6 (with the idea to recapture on d4 with the knight, which I didn't realise) was too passive (??!) and that b5 would look more intimidating (well, it wouldn't have been be such a big mistake had I realised that the doubled pawns were no big deal and shoved the a-pawn down with the support of the f8 bishop, but I got too lost and nervous).

@Rei: I do think that my castling decisions are objectively bad, in the above games in particular. I guess I'm smart enough to avoid castling into a direct attack, and I don't remember any blunderous decisions you're talking about.

Besides, I believe that tactical oversights are consequences of strategic errors - in particular, tactics against the king arise when he's put on the side of the board where opponent's pieces are concentrated or can be switched over relatively fast. Even when I don't miss a direct mating attack, calculating lines deeper in order to avoid it tires me, and so I become prone to missing other tactics too.

19. Nb6 is just a blunder; you were fine until then, and 0-0-0 looked good. The omissions of 13.-a5 and 15.-Bf8 in the other game weren't strategic mistakes or even related to them -- you missed concrete ideas (maybe you miscalculated or didn't have the right move-search heuristics in place, I don't know).

Yeah, technically you played 15.-0-0?! instead of 15.-Bf8!, thus making an "objectively bad castling decision"; but that would be kind of like playing 15.-Rc8?! and calling it an "objectively bad rook development decision" and thinking you need advice on when to develop your rooks.

It would be a displacement of error -- you'd need advice that would lead to playing moves like 15.-Bf8 instead of [something else].

It seems like you don't like to think in terms of variations if you can help it (the whole OP is a sea of basically "X move is [evaluation] because [vague positional ideas]").* That doesn't seem good. After 13.-Bb7 14. a4 0-0 15. axb5 axb5, I'm not seeing how opening of the a-file is to White's advantage.


*Like, here:

Why not just say, "Because after something like 1. Rg1 [a strong black move] 2. Kd2 [a strong black move] 3. [a strong white move] [a strong black move], I have some coordination issues and I'm in a bit of trouble. I would need to insert Bb2 or Be3." In other words, show, don't tell.

Rg1/Kd2 doesn't exist in some strategic vacuum; Black has agency too. And Black's resources (expressed as specific moves) are going to help determine whether Rg1/Kd2 is any good or not -- it's often the case that minor concrete nuances, which you have to account for, will be the deciding factor.

13.-Nc6 is fine, too (it addresses the threat to the d4 pawn).

Now I know the exact move order- about that opening- I don't like 7...e6. I play the black side and 7...Nxb5 followed by d5 is the main equalising line, and 7...a6 is also a good try for black to play for a win imo.

You're right, Nxb5 is the main line, but according to Chesstempo, its results for Black are significantly poorer that in the a6 and e6 lines. However, a big advantage of playing a6 first (before e6) is that, if the bishop retreats to c4, then Black can play d5 with a tempo as soon as White plays e5, and hence have time to avoid the pawn doubling even if White doesn't take en passant. In the Sousa vs Margarido game, Black neglected the Nxd4 threat (created by e5), played 9... Ne7 and ended up drawing (despite following lkasigh's advice about f6); 9... Nxf3+ woulda been the best move, with an edge for Black (-0.5 according to Stockfish 5: 10. Qxf3 Ne7 11. b3 (preventing c4) O-O 12. Ba3 Qc7 13. Na4 b6 14. c3 f6 15. exf6 Rxf6). Thanks for the move order tip!

This line is not that good for white, and usually black players tend to be higher rated.

For example 6...Nbd7 instead of 6...e5 in the mainline King's Indian Classical has (or had - a few years ago in megabase) a higher winrate for black, but this is just because superior black players opted to avoid the drawish dxe5 queenless middlegame to play for a win.

I don't like playing ...Nxb5 either and find it a more difficult position to win after white plays the e5 close position structure and preferred ...a6. But this may be because I played KID as black and also the dragon, so I'm familiar with these structures, whereas I have never played the French or Nimzo or other light-squared strategy openings.

Also in general- against a similar anti-sicilian the KIA, the best line for black is not to play ...e6. The KIA is not that bad against ...e6 sicilians imo, but does poorly against either a normal fianchetto with pawn on ...e7 (and playing ...Rb8, ...b5 plan), or imo the best refutation of playing the c5-d6-e5-g6 structure. Stopping white from playing b4 with a5 (when he plays both c3 and a3), and just either go for ...f5 or ...d5. Black can easily be better out of the opening in this line (in fact this structure is so good that black can even play ...e6 and then ...e5 wasting a tempo).

tbh in the Rg1 example, i'm not really sure why anything other than 0-0 would come to mind unless 0-0 would fail for tactical reasons. there's no queen, so your light squared weakness on the kingside really shouldn't be an issue in terms of castling, but there are still enough pieces on the board that your king isn't going to be happy in the center. and when you play Rg1, that rook now needs three more moves to have any future potential. a king move to get it out of the way, a rook move, then a king move back. that just doesn't really make much logical sense to me.

Well, I wouldn't even call it the 'light-square weakness'; in fact, I had played g3 exactly on the previous move (the prior game moves were Rh8-g8 g2-g3 Nb8-d7), and that was for defensive purposes - I thought that the pawns would be less vulnerable this way. The engine actually liked the g3 move.

It's great that there's a consensus about castling in this spot. I guess that I should just castle by default even in any complex endgame (even queenless, but with, say, both rooks and at least two pairs of minor pieces still on) because connection of the rooks is more important than centralisation of the king - as airwave16 mentioned, the king would still feel vulnerable in the centre.