Caro-Kann Advance 3...Bf5: nine novelties

3...Bf5 is the classical heart of the Caro-Kann Advance: Black develops the bishop outside the pawn chain before playing ...e6, reaching one of the opening's most established structures. White, in turn, has developed a wide range of setups and move orders to put pressure on it.

That makes 3...Bf5 the largest section of our Caro-Kann selection. Using CheckMeta64's novelty recipes, GM Alvar Alonso Rosell picked out nine recent games. They cover all four of White's main tries: the trendy 4.Nd2, the Short System with 4.Nf3 and Be2, the direct 4.h4, and the wild 4.g4.

1. A new plan against the world champion: Pranesh – Gukesh

We start with the trendy 4.Nd2. In Alonso Rosell's words, it's popular because the Nf3-Be2 setup has a tough fight against Black's ...h6 and ...g5 plans, as well as ...a5 ideas. Against Gukesh's solid 6...h6, Pranesh introduced 7.Ne2!?, developing the knight straight to g3 to fight for the kingside light squares.

Gukesh answered with the dubious 9...c5?!, when 9...Nf5 "should be the clean solution". The notes also cover the alternative 6...h5, including Shankland's precise handling against Rapport at the World Cup in Goa.

Takeaway: even the world champion can be caught by a fresh plan in a quiet-looking position. Against 4.Nd2 with h4, remember the clean equalizing route via ...Nf5.

2. A 2148 outprepares a 2420: Agnivo Chakraborty – Illner

Still in 4.Nd2 territory, but now with the aggressive g4 treatment. The position after 5.g4 Bg6 6.Ne2 c5 7.h4 h5 8.Nf4 mirrors the well-known main line with 4.Nc3, included in the notes for comparison: the same plan, with the knight on d2 instead of c3. Here Illner reacted with 8...Bh7, the normal move in the Nc3 version, and White played 9.g5 followed by 10.g6!, and the engine already likes White a lot. The paradox is in the notes: 8...Ne7! is objectively best, but as Alonso Rosell puts it, it performs terribly in practice, and the forcing lines after it are genuinely interesting.

The rating gap tells the story of preparation beating strength: a 2148 dismantled a 2420 in the opening.

Takeaway: a familiar-looking position is not the position you know. 8...Bh7 is the standard move with the knight on c3; with it on d2, it walked straight into 10.g6!. And the objectively best 8...Ne7! performs terribly in practice anyway.

3. A novelty is only as good as your memory: Topalov – Erdogmus

Now the Short System: 4.Nf3 e6 5.Be2. Erdogmus showed the new trend 5...a6!?, a clever idea aimed at lines where White plays Be3 against the ...c5 setups: Black gets the critical position with a7-a6 as an extra tempo.

The way to punish it is 9.g4!, and the key position of that line has so far only appeared in correspondence games. Topalov's 9.f4 was itself a novelty, but not the critical test. Then came the strange part: Erdogmus answered with 9...h5, when 9...Nbc6! was, in Alonso Rosell's words, "the ONLY point of the prep". He forgot the idea? White got an excellent position out of the opening, but the advantage slipped away in the complications and Erdogmus went on to win a wild game.

Takeaway: half-remembered preparation can be worse than none. If 9...Nbc6! is the only point of 5...a6, you'd better know it when the position appears.

4. Same plan, better square: Movahed – Kushko

Same Short System, but with Black's ambitious ...a5-a4-a3 push. White's plan in this structure is the c2-c4 break, and it had already been tried with 11.Nb1: three games went wrong for White there, since ...g5! gives Black good play in the center. Movahed's 11.Na4! is a good conceptual novelty. It prepares the same c2-c4 break, but from a4 the knight also fights against Black's ...c5 counter. As Alonso Rosell puts it in the notes: "Compare with a knight on a4!"

Takeaway: a novelty doesn't need to be a forcing shot. Sometimes it's a known plan executed from a better square.

5. An extra idea from a quiet position: Ayats Llobera – Ohler

One more from the Short System, this time against Black's ...f6 break. Ayats Llobera's 10.Nh4 is what Alonso Rosell calls "a very interesting extra idea": instead of the most played 10.Re1, White immediately targets the f5 bishop before recapturing in the center.

The critical detail is 12...Nfxe5!: only this knight can capture. The notes show that after 12...Ndxe5 White gets a clear advantage with a forcing sequence ending in Nxf5. "Amazing prep", in Alonso Rosell's words. After Black's precise 13...Kf7! White emerged just slightly better, and the game was eventually drawn.

Takeaway: even well-mapped positions leave room for a new approach. 10.Nh4 asks Black a question nobody had asked before, and only one recapture answers it.

6. A model plan for Black against 4.h4: Kollars – Donchenko

Over to 4.h4 h5. Kollars met the bishop trade with 7.Nf3, when 7.Bg5 is the main line, and Donchenko showed that the sideline brings White nothing new. His novelty 9...Rc8 is part of a clean plan: play ...c5 early and fight for the dark squares. By move 14 Black had, in Alonso Rosell's words, "an amazing position".

The rest was what he calls a perfect game from Black: Donchenko kept a small pull all the way and Kollars had to work for the draw.

Takeaway: if you play the Caro-Kann against 4.h4, this is a model equalizing plan: trade the bishops on d3, hit with an early ...c5, and fight for the dark squares.

7. A good novelty isn't enough: Mariot Vaz – Rodi

Same 4.h4 h5 line, and a reminder that finding a novelty and converting it are two different skills. After Black's inaccurate 9...c5?!, White found 11.Qc2!, a good novelty, the thematic unpinning move.

But one move later he went wrong: 12.cxd5? let Black take over, when 12.dxc5! followed by cxd5 and Ne4 would have been winning; the notes show Black's kingside simply doesn't work in that line. Instead, "already Black is better" after 13...cxd4, and Rodi ground out the win.

Takeaway: the novelty gets you the better position; it doesn't play the follow-up for you. One natural-looking recapture threw away everything 11.Qc2! had earned.

8. When the novelty is a strange knight move: Tzagarakis – Megalios

A different flavor of 4.h4: Black met it with 4...Qb6, and after White's inaccurate 5.h5 (5.g4 is known to be good for White), Black uncorked 5...Nh6. A strange move, but Alonso Rosell admits: after it, he likes Black's ideas.

The follow-up 6...f6! is the point, and the notes contain a gem: after 7.Bxh6?! gxh6 8.exf6 Qxb2 9.Kd2, the machine-like 9...Bc8!! keeps everything unclear. Black retreated with 9...Qb6 instead, and White got a clear advantage, though the game eventually ended in a repetition.

Takeaway: a novelty doesn't have to change the plan, only the move that executes it. 5...Nh6 stops White's expansion just like the standard 5...h6, but it develops a piece and drags the game out of known theory.

9. A new idea in the 4.g4 jungle: Bharath – Tan

We teased this one in the series hub, and here's the full story. The wild 4.g4 leads to some of the densest theory in the Caro-Kann, and after 4...Be4 5.f3 Bg6 6.e6 fxe6 7.h4 h5, Bharath's 8.Nh3 is exactly the kind of novelty that works in practice: a new idea in a dense line.

Objectively? Many moves work for Black. Alonso Rosell's notes show that in the 8...hxg4 line Black "can choose to play ANY move" and there's no difficult idea to find. Except the one Tan played: 8...Nf6?, and White's initiative rolled from there.

Takeaway: in dense, forcing theory, a novelty's practical value can be enormous even when the engine shrugs. Your opponent has to solve a brand-new problem with the clock running.

That's the 3...Bf5 material: nine novelties, four different White systems, and the same lesson in every one of them. The classical main line of the Caro-Kann Advance is anything but settled theory.

All nine games came out of a single search: the Caro-Kann starting position, filtered through the novelty explorer. If you play 3...Bf5 with either color, you can run the same search and see what turns up in your own lines. Part 3 (the sharp 3...c5) and Part 4 (the sidelines: 3.f3, the Exchange and 2.Nc3) are coming next.