Where middlegame plans come from when you know none

The short answer

A plan comes from three questions, asked in this order: what is my worst piece, where is the opponent's fixed weakness, and which pawn break does my structure allow. The answers give you the three plans that cover most amateur middlegames — improve the worst piece, pile onto the weakness, play the break.

You do not need to have memorised anything. Every one of those questions is answered by looking at the board for ninety seconds. What follows is how to ask them properly, and one worked structure for each.

The pawn structure tells you which side of the board to play on

Pieces move. Pawns mostly do not. So the pawns are the part of the position that will still look the same in fifteen moves, which makes them the only reliable thing to base a plan on.

The rule of thumb is old and it holds up: play where your pawn chain points. If your pawns run c3–d4–e5, they point at the kingside, so your pieces belong there. If your pawns run f2–e3–d4 against a black chain on e6–d5, you have a queenside majority and the play is there.

This is why "what is the plan" is the wrong first question and "what does the structure look like" is the right one. In a Carlsbad structure the plan is a minority attack whether you are 1400 or 2400. In a Benoni structure it is e4–e5 or f4–f5. You are not inventing anything. You are recognising a shape.

Plan 1: improve your worst piece

Look at your pieces and name the one doing the least. Then find it a job. This is the plan you use when nothing else is obvious, and "nothing else is obvious" describes a lot of positions.

Concrete case. You have the Italian with pawns on c3 and d3, bishop on c2, knight on f3. Your knight on b1 is the worst piece on the board — it has one square and no future. The standard route is Nbd2–f1–g3, and from g3 it hits f5 and h5 and supports a kingside build-up. Three tempi to turn a dead piece into your best attacker. That is a complete plan and it required no memorisation, only the observation that the knight was doing nothing.

Rooks count too. A rook on a1 behind an unmoved a-pawn is not a piece, it is furniture. If you cannot find a break and cannot find a weakness, put your rooks on the files most likely to open and wait.

Plan 2: find the fixed weakness and pile on

A weakness you can attack has to be fixed — a pawn that cannot advance out of trouble, or a square no pawn can ever cover. Backward pawns on half-open files, isolated pawns, doubled pawns and holes in front of them all qualify. A pawn that can just push forward next move does not.

The minority attack is the cleanest example. Exchange Queen's Gambit: White has pawns on a2, b2, d4, e3; Black has a7, b7, c6, d5. White plays b4, a4, b5, and after bxc6 bxc6 Black is left with a backward pawn on c6 on a half-open file, plus a weak c5 square. White then triples nothing and invents nothing — rooks to c1, knight to a4 or e5, and the pawn is a permanent target for the next twenty moves.

The order matters: fix first, attack second. If you attack c6 while Black can still play c5, you are attacking air.

Plan 3: play the break

Closed and semi-closed positions do not open themselves. If your pieces are better placed than your opponent's, you need a pawn break to give them something to do, and in most structures there are only one or two legal candidates.

King's Indian with White pawns on c4, d5, e4 and Black on d6, e5: Black's break is f7–f5, White's is c4–c5. That is it. Those two moves are the whole strategic argument of the opening. French Advance, White e5 against Black e6–d5: Black breaks with c5 and f6, White holds with c3 and f4. Again, two candidates each.

The practical version of this plan is a preparation list. Before you push, ask what you need in place first — the rook behind the pawn, the king off the file that is about to open, the piece that currently blocks the break moved out of the way. Most failed breaks fail because they were played one move before the preparation was finished.

Planning is not usually what costs you the game

Here is the uncomfortable part. Across 275,263 engine-classified moves from 8,640 games analysed on Mated, 3.7% of moves played are outright blunders and another 4.1% are mistakes. That is roughly one real error every 13 moves, before you count the 18.3% that are inaccuracies. No plan survives that rate.

The same data set shows where it happens: 8.9% of middlegame moves are a mistake or worse, against 2.3% of endgame moves. The middlegame is where the errors cluster, which is exactly the phase people want plans for. It is tempting to read that as "I need better plans". Read the next number first.

The average move in that data gives away 92 centipawns against the engine's best, but the median is far lower — a small number of very bad moves carry most of the damage. Your average middlegame move is not the problem. Four moves a game are. Those four are almost never failures of planning; they are failures of checking what your opponent's last move attacked.

So the honest version of the advice: a plan is what you do with the 90% of moves that are fine anyway. It sets the direction. It does not protect you from the loose bishop on b4. Both need work, and if you had to pick one, pick the blunders — they are worth more centipawns and they are easier to reduce.

How to tell whether you actually had a plan

Open one of your recent losses. At move 15, before looking at what you played, write one sentence: what were you trying to do. If you cannot write it, you were not planning, you were responding.

Then play through moves 15 to 30 and mark every move that served the sentence. In my experience of doing this with club players, the count is often two or three out of fifteen. The rest are reactions, repositions, and moves played because it was your turn. Fixing that is not about learning more plans. It is about noticing, at the board, when you have stopped following the one you had.

The same review will show you which of the three questions you skipped. Games where your rook never left a1 are worst-piece games. Games where you shuffled for twenty moves in a closed position are break games. Tools help with the counting — Mated runs Stockfish over every position of your last games in the browser and scores the phases separately, so you can see whether your centipawn loss concentrates in the middlegame or somewhere else — but a board and an honest fifteen minutes gets you most of the way.

A note on how hard your training should feel

One number from our puzzle data is worth thinking about even though it is not about planning. Across 2,296 rated puzzle attempts on Mated, 94.2% are solved first time. That is a very high pass rate, and it means a lot of those attempts taught nobody anything.

Apply the same test to strategy work. If you look at a middlegame position, say "minority attack", and you are right, you have confirmed something you already knew. The positions worth studying are the ones where you produce a plan, check the master game, and find they did something else entirely — then have to work out why. Those are uncomfortable and there are fewer of them, which is precisely the point.

Nobody has measured well what the ideal failure rate is for chess study. But a set of positions you pass 94% of the time is a test, not a lesson, and you should know which one you are doing.