A kingside attack does not start with a sacrifice. It starts with a question: if my opponent had the move, what would they do with it? In Buenos Aires in 1914, playing two opponents in consultation, Capablanca had the bishop pair, a pawn wedge on f5 and every reason to rush at the king. Instead he stopped to take away Black's counterplay, first with 12.a3 and then with 13.Be6, and only then brought his pieces across. When he finally sacrificed with 23.Nxh7 he had counted his attackers, and he followed up with the one move, 25.Bxc4, that kept the attack alive. Then, with Black tied up, he opened a second front with 30.b3 and made a passed pawn on the far wing. Take one idea from this game: before you attack, ask what your opponent wants, stop it with a move that also helps your plan, and count your attackers before you give up a piece.
Every move, ask what your opponent would do with the move. Then stop it with a move that also improves your own position. That is what I call proactive prevention.
Bring three pieces to the king before you give up material. Then look for three candidate moves, because the attack can still slip away after the sacrifice.
When the opponent is tied to one wing, make a new problem on the other. Capablanca's passed a-pawn stretched a defense that was already stuck.
Watch the lesson, then play the game through on the board below. The board stays with you as you scroll. Make each move yourself while you read. You'll remember a plan you played with your own hands far longer than one you only watched.
The first thing I want you to notice is what Capablanca does with a pleasant position: he does not simply go after the king. At move 12 he sees that Black wants ...Na5 to remove his light-squared bishop, the piece holding back ...d5. So he plays 12.a3, giving the bishop a safe retreat to a2. When Black prepares ...d5 with 12...Be8, he answers 13.Be6, so that ...d5 can be met by exd5 and d4, and the center breaks open in his favor. This matters because an attack on the wing rarely works while the center is unstable. Only after those two moves does he swing the queen to h4 and push g4 and g5. When the sacrifice 23.Nxh7 finally comes he has three attackers ready, knight, queen and rook, and the engine agrees it is the right move. It is sound, but it is not a forced win: with best defense White is ahead by a little over a pawn's worth, which is why the next decision needed real care. Then comes the last idea: with Black's pieces stuck on the kingside, 30.b3 starts a passed pawn on the other wing. Prevention, then calculation, then a second front.
Capablanca left St. Petersburg in July 1914 and sailed for Buenos Aires. In his own words, the war "broke out as I was in mid-ocean", but he arrived safely and played what he calls "several games of the so-called brilliant kind", crediting the inexperience of his opponents for the chances. This one, played on September 17 against L. Molina and E. Ruiz in consultation, opens 1.f4 e5, the From's Gambit, where Black gives a pawn for a fast attack on the white king. The engine would take the pawn with 2.fxe5, and it rates 2.e4 as slightly better for Black. Capablanca plays 2.e4 instead, and I agree with him. If you have a choice between taking material and defending, or giving material and attacking, choose the attack. After 2...Bc5 we have a King's Gambit Declined. Then 5.Na4 and 7.Nxb6 win the bishop pair, and 10.f5 builds a wedge that threatens f6. Black stops that with 10...f6, the bishop checks on c4, and after 11...Kh8 the lesson begins.
Look at White's position after 11...Kh8. The king on h8 has no squares, the bishop on c4 points at it, the knight can jump into the kingside, and the queen and rook can follow. Everything says attack. Material is level: White has the bishop pair and Black has an extra knight in place of a bishop. The position is better for White, but only modestly, about half a pawn on the engine. Before you choose an attacking move, do what Capablanca does here. Ask the question that changes how you play: if Black had the move, what would they do with it?
White has the bishop pair and a wedge on f5. If it were Black's move, what would they play, and how would you stop it while keeping your attack?
Look at the bishop on c4. Which black knight can challenge it, and which pawn break is it holding back? Find the threat first, then the move that answers it. Think, then reveal.
Follow the order of Capablanca's moves, because the order is the lesson. 12.a3 answers ...Na5. After 12...Be8, the point of which is ...d5, comes 13.Be6: in his note he says it is "to prepare against Black's d5", which would now be answered by exd5 followed by d4, and the engine has both moves as its first choice. People call this prophylaxis. I do not like the word, because it sounds passive. Proactive prevention means you stop their plan and improve your own at the same time. Only now does the attack start: 14.Qe1 and 15.Qh4, then 16.Ba2 to keep the bishop. At move 17 he makes a very human choice. 17.Bxf7 was technically better than 17.c4, by about half a pawn, but he keeps pieces on the board, because the more pieces your opponent has to handle, the more mistakes they make. I am not fully sold on it, but understand the thinking. Then 18.g4, 19.Bd2, 20.g5 and 22.Rf3. The knight, queen and rook are all aimed at h7. That is the rule of three from my Logical Chess series: three pieces at the enemy king before you commit.
It is White to move. Black has just taken a pawn with 22...bxc4 and is ahead on the board. The knight on g5, the queen on h4 and the rook on f3 are all pointing at the king, and there is a natural sacrifice in the air. The question is not whether it looks attractive. The question is whether it works. Count the attackers, look at what each defender can really do, and find your three candidates before you decide.
White to play. Is the sacrifice on h7 sound? If you play it, what is your next move, and what must you see before you let go of the knight?
23.Nxh7, and it is sound. After 23...Nxh7 24.Rh3 the knight on h7 is pinned twice over and Qxh7 is mate, so Black must defend at once. If 24...Bh5, White has 25.Qxh5, and Capablanca notes that 25.Bxc4 might be stronger still; the engine agrees. So 24...Bg8 is the move. Capablanca wrote that "a very careful analysis will demonstrate that the sacrifice is absolutely sound", and a century later the engine has 23.Nxh7 as its first choice. He also rejected 23.Rh3, because of 23...h6 24.Nxf7+ Nxf7 25.Bxc4 d5, and that line does run to dead level, so his judgment holds up. The rook move is not hopeless, since 24.Bxc4 instead keeps a small edge, but the sacrifice is better. Now the warning. Sound does not mean winning by force. After 24...Bg8 you have given a piece, and only one move keeps the advantage: 25.Bxc4. It brings a third attacker in and picks up a pawn. Every other move leaves White worse, by about two pawns. This is where players throw games away. The attack feels easy, the natural moves come quickly, and you stop looking. Black's best was 25...d5 26.Bxd5, when White has two extra pawns and a raging attack for the knight, roughly a pawn better on the engine. In your games, once you have sacrificed, slow down. Three candidates, rank them good, better, best. That is the Goldilocks principle, and it matters most when you feel most confident.
Black's natural defense was 25...Nf7, but then 26.f6 is a brilliant resource: if 26...gxf6 then 27.Qxf6 is mate, because the knight on h7 is pinned by the rook and cannot take. In the game they tried 25...Rf7. On the video I said you should not take the rook, because it is a bad defender. That was Capablanca's choice with 26.Kh1, and it keeps every piece on, but to be accurate the engine slightly prefers winning the exchange with 26.Bxf7. Both are good, and White is well ahead either way. Now every white piece joins in: 27.Bd5, 28.Rg1, 29.Bg5. Then comes the position in the diagram, with Black's pieces stuck and no counterplay. The direct 30.Bxf6 was stronger by about two pawns, and if you chose it, full marks. Capablanca played 30.b3, because "White proceeds to obtain a passed Pawn with which to win the game". That is the principle of two weaknesses in a middlegame. When the opponent is tied to one wing, you open a new front on the other, and 31.a4, 33.a5 and 34.a6 walk the pawn down the board.
Capablanca is not perfect in the finish. At move 35 the capture 35.Bxf7 was stronger than 35.Bxf6 by about two pawns, and I think that is a small sign of something bigger: he was exceptionally good at finding good moves, and less driven to find only moves. Here it costs nothing. After 35...Nxd5 he checks first with 36.Bxg7+, and 36...Rxg7 37.Rxg7 Kxg7 38.Qh6+ Kh8 39.Qxd6 leaves one of the strangest positions you will ever see. Black has a bishop and two knights against White's rook, and is helpless. Capablanca put it well: "with two pieces ahead Black is helpless". The knight on h7 is pinned, and if the knight on d5 moves then Qf6 is mate, but if it stays it can be taken with the a-pawn still running. Capablanca records that Black resigned a few moves later, though neither the book nor the study gives those moves. Notice also his note to 38...Kf7, where after 39.Rg3 he trades the queens and rook to reach a won ending with his passed pawns. Even with a king to hunt, he thinks about the endgame.
The principle: for half the game it is not your move, so prepare for what your opponent will do. In this game: 12.a3 and 13.Be6 took away ...Na5 and ...d5 before the attack began. In your games: before each move, ask what your opponent would play if it were their turn.
The principle: if the center is unstable, an attack on the wing generally will not work. In this game: Black kept trying for ...d5, and Capablanca kept an answer ready. In your games: look at the center before you push pawns at the king.
The principle: two pieces rarely break a king's defense, so bring a third. In this game: knight, queen and rook stood ready before 23.Nxh7, and the bishop came next. In your games: count attackers and defenders, then calculate. Nothing replaces calculation.
The principle: rank three candidates good, better, best, the Goldilocks principle. In this game: 25.Bxc4 was the only move that kept the advantage after the sacrifice. In your games: in a long game, stop and compare, especially when your attack feels easy.
The principle: one weakness can often be defended, two usually cannot. In this game: with Black tied to the kingside, 30.b3 started a passed a-pawn. In your games: when your opponent is stuck on one wing, look for a new target on the other.
You start by making sure the attack can work. Check three things. Is the center stable, or can your opponent strike back there? Which of your pieces can reach the king, and are there at least three? And what is your opponent planning if you ignore them? Capablanca had the bishop pair and a wedge on f5 at move 12, yet his next two moves dealt with ...Na5 and ...d5. Only then did the queen go to h4 and the g-pawn start to move. Pawn storms and sacrifices come last, once the pieces are in place and the counterplay is gone.
Prophylaxis simply means a preventive measure, and I think the word sounds too passive for what strong players do. Proactive prevention is an active habit. You ask what your opponent wants to do, and you look for a move that stops it and also improves your own position. Here 12.a3 stops ...Na5 from winning the bishop and also prepares the retreat to a2, where the bishop stays aimed at the king. 13.Be6 prepares the answer to ...d5 and plants the bishop deep in Black's position. Neither move slows the attack down.
Yes, and the engine agrees with his note: 23.Nxh7 is its first choice. But be precise about what sound means. It does not mean a forced win. After 23...Nxh7 24.Rh3 Bg8 25.Bxc4 d5 26.Bxd5, White has two pawns and a strong attack for the knight and stands about a pawn better. The large advantage came later, because Black chose 25...Rf7 instead of 25...d5, and then could not free the position. Capablanca also needed the exact follow-up. A sacrifice that needs one only move right after it is sound only if you have seen that move.
Because he was following a plan he trusted, and he had no need to hurry. The engine has 30.Bxf6 ahead by about two pawns, and I think the fair verdict is that both moves win easily. So if you chose the capture, you found the better move. What 30.b3 shows is how a strong player thinks when the opponent has no counterplay. He stretches the game by making a second problem, a passed pawn on the other wing, so that a defense already tied to the king now has to stop a pawn as well. Keep the plan in your head, and still check whether something more direct is available.
Playing only your own moves. Most attacking players look at the enemy king and forget that the person across the board has ideas too. For half the game it is not your move, and you have no control over what happens next unless you have prepared for it. Train yourself to ask, after every move your opponent makes: why did they play that, and what would they do next if I passed? In this game that one question produced 12.a3 and 13.Be6, the quiet moves that made the attack possible. Add it to your games and your attacks will stop falling apart.
Six kingside attacks, rising in difficulty, each one ending in mate. Count the attackers, check what the defenders can really do, and look for the checks and captures first. Drag the pieces; the board answers for the defense.
Knowing how to attack is one thing. Knowing when your attack is ready, and what your opponent is planning while you build it, is what decides your games. In a 2-hour Strategy Assessment I go through your games personally and build you a roadmap for your chess, around the habits that are really costing you points.
Book Your $130 Strategy Assessment 2 hours via Zoom · Worldwide · Limited spots