Strategy — Where Plans Come From 因勢利導
Almost nobody is stuck because they lack a plan. They are stuck because they think a plan is something you invent. It is not. In the overwhelming majority of positions you will ever reach, the pawn structure already contains the plan, and your job is to read it off. That is what this page teaches, and it is the whole of it.
When strategy starts to matter — and why that claim is contested
This site's line is that strategy is close to irrelevant below about 1150 — where the Ladder first puts plan-making, at Rung 3 — and then suddenly is not: below that, games end when someone drops a piece, and a better plan does not survive a hanging knight. That is a judgement, not a measurement, and the Ladder states it as a sequencing decision rather than as a finding. No dataset has ever classified errors as "tactical" or "positional" by rating band, because engines do not produce that category.
And there is a real argument on the other side. FM Nate Solon points out that the usual proof — run amateur games through an engine, observe that the big swings are tactical — is definitionally rigged: "The moves that produce the biggest evaluation swings are, almost by definition, tactics. But in a practical game there is great value in giving your opponent a narrower or more difficult path to keep the balance, even if the engine can see that they are still surviving with perfect moves." Strategy in his account is the jab, not the knockout — invisible in the highlight reel, and what sets the knockout up.
He is right that "the data proves strategy doesn't matter" is a measurement artefact. We still order the curriculum the same way, for a reason about attention rather than truth: a player hanging a piece every third game has exactly one job (Rung 1), and this is not it. If you are above about 1150 and still treating strategy as optional, you are who this page is for.
Pawn structure is the plan generator
Pawns are the only pieces that cannot move backwards. Every other feature of a position is temporary; the structure is what persists, and it decides which files will open, which squares can never be defended, and which wing you belong on. That is why structure — not opening names — is the right unit of strategic knowledge. GM Mauricio Flores Rios's Chess Structures: A Grandmaster Guide (Quality Chess, 2015) covers the 28 most common structures in 140 games and fragments; the seven below are the ones you meet between 1150 and 1900.
For each, the test is not "can you name it." It is can you state, in one sentence each, what it tells you to do and what it tells your opponent to do. A structure that only tells you your own plan has not been understood.
1 · The isolated queen's pawn (IQP)
One side has a d-pawn with no pawns on the c- or e-files. It arises constantly: the Panov–Botvinnik Caro-Kann 1.e4 c6 2.d4 d5 3.exd5 cxd5 4.c4 Nf6 5.Nc3 e6 6.Nf3 Be7 7.cxd5 Nxd5 gives White one on d4; the Tarrasch and Semi-Tarrasch give Black one on d5.Isolated queen's pawn · White's d4, with c5 and e5 as the squares it buys
If you have it
Keep pieces on. The pawn buys you the c5 and e5 squares, a space edge
and two half-open files. Those squares are not outposts in the strict sense used below — Black can
challenge them with ...b6 or ...f6 — but both advances loosen him further, which
is the point. Aim at the king, play for the d4-d5 break, and decline every
queen trade you reasonably can. The asset is dynamic and it expires.
If you are against it
Trade pieces, blockade, then attack it. A knight directly in front of the pawn is the classic blockader — no pawn can dislodge it and it still radiates. Trade minors, keep a rook on the d-file, head for an endgame where the pawn is a static weakness with nothing left to compensate.
The one-line version: the IQP is strong in the middlegame and weak in the endgame. Every exchange moves the position from the first state to the second, which is why the whole struggle is about trades.
2 · Hanging pawns
Two friendly pawns abreast — classically Black on c5 and d5 — with no pawns on the adjacent b- and e-files, so the opponent has two half-open files bearing down on them. They arise from the QGD Tartakower, the Queen's Gambit Accepted and several Nimzo-Indian lines.Hanging pawns · Black's c5 and d5, White's rooks on the half-open c- and d-files
If you have them: they are a mobile centre until they stop moving. Play for one of the two advances
— ...d4 or ...c4 — at a moment when it gains something concrete, and use the space
in front of them for a rook lift. Never advance one merely to relieve the pressure.
If you are against them: pile up on both in a way that provokes an advance. The moment one advances the pair splits — the advanced pawn can be blockaded, the one left behind becomes backward or isolated on a half-open file. A fixed target instead of a moving one. Hanging pawns are strong while mobile and weak once fixed, and the whole struggle is over that transition.
3 · The backward pawn and the hole behind it
A pawn on a half-open file that has fallen behind its neighbours and can no longer be defended by a pawn. The canonical case is Black's d6 pawn in the Najdorf after 1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be2 e5 7.Nb3 — the c-pawn is gone, so only pieces can ever defend d6, and White's rooks belong on d1.Backward d6 pawn · and the permanent hole on d5
The second half matters more. Behind every backward pawn sits a hole: a square no enemy pawn can ever control. Here it is d5 — Black's c-pawn is gone and the e-pawn is already past it, so d5 belongs to White for the rest of the game.
What you do differently: if you have the backward pawn, your entire plan is
to get rid of it — engineer the freeing advance (...d5 here) even at the cost of a pawn. If you
are against it, do not rush to win the pawn. Occupy the hole with a minor piece, double on the file,
and force the defenders into passivity. The pawn falls later, or the position collapses elsewhere while
everything is tied to it.
4 · Doubled pawns
The most over-punished structure in amateur chess. Doubled pawns are a liability only when they are also immobile and on a half-open file. Otherwise they come with compensation you paid for: an extra half-open file, extra control of the squares beside them, and usually the bishop pair. The Nimzo-Indian Sämisch hands White doubled c-pawns for the bishop pair; the Exchange Ruy Lopez 1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Bxc6 dxc6 hands Black doubled c-pawns for the same thing.
The Exchange Ruy is the best structure lesson on this page, because the plan is unusually explicit. After 5.O-O f6 6.d4 exd4 7.Nxd4 White has a healthy four-against-three kingside majority; Black has a four-against-three queenside majority that is crippled — four pawns on three files, which cannot force a passed pawn against correct play. So White's plan is stated for him: trade absolutely everything and head for the king-and-pawn endgame.Exchange Ruy · White's healthy 4-v-3 kingside, Black's crippled 4-v-3 queenside That endgame is a structural plus, not a forced win — Black's bishop pair and the tempo count are real compensation, and top-level practice treats the line as balanced. Black's plan is the exact negative: keep pieces on, use the bishop pair, and never let the position simplify. Two players, one structure, opposite instructions.
5 · The Carlsbad structure and the minority attack
White pawns on a2, b2, d4, e3, f2, g2, h2; Black on a7, b7, c6, d5, f7, g7, h7. White has traded the
c-pawn for Black's e-pawn, so White has a half-open c-file and Black a half-open e-file. It comes from
the Queen's Gambit Declined Exchange — 1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.cxd5 exd5 5.Bg5 c6 6.e3
— and was named for the 1923 Carlsbad tournament, where the formation's popularity took off. (That move order
is the one the repertoire plays, which keeps ...Bf5 available; the
structure is the same whichever way you get there.)Carlsbad structure · White's minority attack runs b4-a4-b5xc6
Worth knowing best of the seven, because it has three named plans for White and they are mutually exclusive. Choosing between them is strategy.
| White's plan | How it works |
|---|---|
| The minority attack | Advance b4, a4, b5 — two pawns against three — and trade on c6. Black recaptures with the b-pawn and is left with a backward c6 pawn on a half-open file: that is the standing target, and it is the whole point of the plan. Black's a-pawn is left isolated as well, but be honest about it — so is White's, the a-file is blocked by both, and the a-pawn only becomes a real second front once a rook reaches the open b-file. Three pawn moves (usually four, counting Rb1) that develop nothing, in exchange for a permanent structural gain. |
| The central break | Support e3-e4 with f3 and Nge2. Trades in the centre and plays for the initiative rather than structure. Chess.com's Carlsbad article gives Kasparov–Short 1993 as the model game; we have not checked that game ourselves, so take the example, not the citation. |
| The Pillsbury plan | Knight to e5, supported by f2-f4, and attack the king. The oldest of the three. |
Black's counter-plans mirror them. Against the minority attack the standard
antidote is to play ...c6-c5 first, accepting an isolated d-pawn in exchange for freedom and an
open position — a worse structure for a better game, which is the trade strategy is made of. The other
counter, and the one this site's repertoire actually recommends: seize e4 with a
knight, put a rook on the half-open e-file, and attack on the kingside, where White has committed no pawns and
every black piece naturally points.
6 · Pawn majorities
Wherever the structure is asymmetric, each side has a majority somewhere; a healthy one is a future passed pawn, a crippled one is a majority in name only. Count them at move fifteen, not move fifty. A queenside majority with both kings castled kingside is worth most, because the passed pawn it produces is far from both kings. Two board actions: advance with the candidate — the pawn facing no opposing pawn — first, and be willing to enter an endgame that looks dead level, because healthy against crippled is not level.
7 · Closed centres and open centres
The highest-level structural question, and the one improvers get wrong most often, because it governs whether the other rules even apply.
Open or fluid centre: play in the centre, develop fast, get the king safe, do not start a wing attack. Rooks belong on central files.
Closed centre — pawns interlocked, as in the King's Indian after
1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Nf3 O-O 6.Be2 e5 7.O-O Nc6 8.d5 Ne7: now wing
play is correct, and you attack on the side your pawn chain points toward. White's chain c4–d5–e4 points
queenside, so White plays c4-c5 at the base on d6. Black's chain e5–d6 points kingside, so Black
plays ...f7-f5 at e4. Both are right. The race is the game.King's Indian closed centre · White plays c4-c5, Black plays ...f7-f5
The board action: before committing to any wing plan, ask one question — can the centre be opened in the next three moves? If yes, you may not attack on the wing yet. If no, you may.
The drill that makes this stick
Take twenty of your own games. At move fifteen of each, write two sentences: the structure's name if it has one, and what it instructs each side to do. No moves, no evaluation. Forty minutes total. That converts this page from reading into Principle Ⅱ retrieval, which is the difference between recognising the Carlsbad in a book and recognising it on your own board with your own clock running.
Piece activity
Structure tells you where to play. Activity tells you which pieces to play with, and the useful list is shorter than most books suggest.
Good and bad bishops
A bishop is "bad" when your own pawns sit on its colour, blocking it. The action is not to mourn it — it is to either trade it or move the pawns. Those are the only two options and improvers take neither.
The qualification is real: a bad bishop defending your fixed pawns is doing a job, and trading it can leave them undefendable. The maxim "bad bishops protect good pawns" is universally attributed to GM Mihai Suba's Dynamic Chess Strategy (1991); we could not verify the wording against the text, so treat the attribution as unconfirmed and the idea as sound. Which minor piece you actually want once the queens are off is a separate question, answered on Endgames.
Knight outposts
A square in or near enemy territory that no enemy pawn can ever attack, ideally defended by one of your own. The authorities disagree on the definition: Nimzowitsch required the square to be on an open file; Hooper and Whyld's dictionary makes the file merely desirable.
Romanovsky's test is the one to carry into a game: a square is weak only if an enemy piece can occupy it and cannot then be driven away or exchanged off. Before spending four moves manoeuvring a knight to a hole, check that they cannot simply trade it.Knight outpost on c5 · it passes Romanovsky's test
Citation warning, same standard as the Calculation page: all three definitions come to us from one secondary compilation that quotes the primary texts with page numbers (My System 32; Hooper and Whyld 285; Romanovsky, tr. Adams, 37). We have not opened those books.
Rooks, open and half-open files
A rook on a fully open file has no target on it; a rook on a half-open file points at an enemy pawn that cannot run. Our claim — a judgement, not a law — is that improvers systematically overvalue the open file and undervalue the half-open one. The open file is the better invasion route; the half-open file is the better pressure point, and it is created by the pawn trade you were probably trying to avoid.
The seventh rank
With the back rank defended, two rooks on the seventh almost never mate by themselves. Their value is attacking pawns, cutting off the king and paralysing the defence — a transitional advantage that must be converted into a material or structural one. Doubling on the seventh and waiting for a mate that never comes is a real way to lose a won game.
The principle of two weaknesses
A single weakness can usually be defended indefinitely, because the defenders only have to cover one place. Advantages convert when a second front opens. It is why the minority attack is not finished when the backward c-pawn appears — the c-pawn ties the defenders down, and the game is decided by what White then opens elsewhere. The instruction: when your pressure on one weakness stops making progress, stop hammering and go make a second one. Note the level, though: Silman files this under Master, and Endgames parks it for the same reason. Recognise it in your own games before you go looking for it.
Weak squares and colour complexes
A hole is one square no enemy pawn can control. A colour complex weakness is several of them on
the same colour with the bishop that covered them gone. The commonest way an improver acquires one is by trading
a fianchettoed bishop: after ...g6 and ...Bg7, the squares f6, h6 and g7 are all dark,
the bishop stood on one of them and was the only piece covering the other two, and ...g6 permanently
removed the pawn that could have covered f6 and h6 itself. Trade the bishop off and all three are soft for good.
Three board actions. Before agreeing to any minor-piece trade, ask which colour squares the departing piece was covering and whether a pawn can cover them instead. When you find an enemy colour complex, keep the bishop of that colour even at material cost — nothing of the other colour substitutes for it. And when the weakness is yours, the repair is almost always a pawn move, not a piece move.
Space, and the trap of taking it
More space means more squares for your pieces and, crucially, the ability to switch wings faster than the opponent can respond. That is the mechanism, not a vague feeling of comfort.
The best evidence anyone has produced: Solon took 100,000 Lichess games (blitz and slower, players within 200 points, material equal after move 15) and measured results against Stockfish's space term. More space was an advantage at every rating level — and the advantage did not grow with rating; if anything it was slightly larger at lower levels. Caveat at the point of use: one blog analysis, one platform, unreplicated, and the article publishes a chart rather than numbers.
The trap
Space is territory you have to defend, and every advanced pawn is one that cannot come back. The mechanism is real — a squeeze means controlling the position for many moves without slipping, and "for a human, the extra blunder risk of having more pieces exposed for many moves starts to add up." But be clear about the status of that sentence: it is the hypothesis Solon set out to test, and his own data did not support it — space still measured as an advantage at the lowest rating tiers. So the caution below is our coaching judgement standing against the one measurement anyone has made, not an extension of it. Do not grab space you have no follow-up for. The test before any space-gaining pawn move: name the square it permanently weakens, and name the piece that will use the space. If you cannot do both, play a piece move instead.
The plan-generation procedure
Here is the repeatable answer to "what do I do here." It is built on GM Jacob Aagaard's three questions from Grandmaster Preparation: Positional Play — where are the weaknesses, which is my worst-placed piece, what is my opponent's idea — with the structural step in front, because at improver level the structure usually answers the first question for you. Same citation caveat as the Calculation page: the three questions reach us through a secondary summary, not the book.
Honest note: nothing like this procedure has been experimentally validated. The one direct test of a structured thinking method we could find — Chessable's 2024 candidate-moves study, 207 players — found no significant difference, and Calculation opens with the same result for the same reason. Treat the five steps as a scaffold against drifting, not as a mechanism proven to produce better moves.
When to trade
Exchanges are where strategy becomes a concrete decision, and the fastest source of rating on this page, because the rules are short and the decision recurs every few moves.
| Situation | Rule | Why |
|---|---|---|
| Ahead in material | Trade pieces, not pawns | Every trade makes the extra unit a larger fraction of what is left; keeping pawns preserves the endgame you are heading for. |
| Behind in material | Trade pawns, not pieces | The exact inverse. Fewer pawns means more drawn endings. |
| Cramped | Trade anything you can | Fewer pieces need fewer squares. The standard remedy for a space disadvantage, and it works. |
| More space | Avoid trades | Space is only worth something while there are pieces to use it. |
| Any position | Trade your bad pieces for their good ones | The most under-used rule here. Before every trade, ask which of the two was doing more work. |
| Attacking | Keep your attackers, trade their defenders | Count the defenders of their king before and after. Most failed attacks failed at an exchange three moves earlier. |
The first rule is universally attributed to Capablanca. We looked in Chess Fundamentals and could not find it stated in those words; the rule is sound, and the attribution is folklore until someone produces the page.
The one genuinely empirical part of this section
IM Larry Kaufman's statistical study of roughly 300,000 master games (2300+) gives numbers you can use at the moment of a trade:
- Knight and bishop are equal at 3¼ — an unpaired bishop is not worth more than a knight.
- The bishop pair is worth half a pawn. That is the figure behind every "doubled pawns for the two bishops" bargain above, and it is why those bargains are usually fair.
- Knights gain value with more pawns on the board; rooks gain value with fewer. Kaufman's adjustment: raise the knight by 1/16 and lower the rook by 1/8 for each pawn above five that your side holds, and reverse both for each pawn short of five.
- The exchange averages 1¾ pawns — and it moves in the direction most improvers guess wrong. With no minor pieces traded it is worth only about 1½; with the queens and a pair of rooks gone it rises to slightly more than 2, falling back to about 1½ if the minor side holds the bishop pair. Each pawn trade helps the rook. Giving up the exchange is dearer in a simplified position, not cheaper.
Kaufman's numbers are averages over a database, not evaluations of your position. They tell you which way an imbalance drifts as pieces come off; they do not tell you whether this particular exchange sacrifice works.
Steinitz, and what the last century did to him
Steinitz's theory as usually taught: the initial position is in equilibrium; advantages arise only from an opponent's error; you accumulate small ones until they amount to enough; and once you have enough, you are obliged to attack, or the advantage dissipates. 不可勝在己，可勝在敵 is the identical thought twenty-three centuries earlier: your own soundness is yours to arrange, the winning moment is theirs to provide.
Caveat in the spirit of Sources: the pithy formulations of "Steinitz's theory" in circulation — including the attack-or-lose-it clause — come from later writers systematising him, not from a quotable line in the Modern Chess Instructor. The theory is real; the wording is secondary literature.
The modern qualification, and what it means for you
John Watson's Secrets of Modern Chess Strategy argues that top-level play since Nimzowitsch is characterised by rule-independence: grandmasters routinely accept structural damage, cede the centre and leave kings uncastled, because concrete dynamic factors outweigh the classical elements more often than Steinitz's system allows.
That does not make the rules wrong. It makes them defaults with named exceptions, and the exceptions always concrete. Follow the structural plan unless you can calculate a specific line that says otherwise. "I felt the position wanted something dynamic" is not a reason. "After this sacrifice I have checked the four defences and they all lose material" is. Rule-independence is earned by calculation; it is not permission to freestyle.
What strategy books get wrong for improvers
Two problems, and they are not the ones the reviews mention.
They teach recognition of positions you will rarely reach. A book of grandmaster games shows structures handled with a precision that only exists when both sides play well, in openings you do not play, in positions your opponents will never steer you into. The recognition is real; the transfer is thin.
They are read passively. The larger failure, and entirely fixable. Reading an annotated game with the moves visible produces exactly the fluency-mistaken-for-mastery of Principle Ⅳ: everything makes sense, nothing is retained.
And the deeper objection, which deserves a hearing
IM Willy Hendriks, in Move First, Think Later, argues the whole genre is post-hoc rationalisation: "No chess player thinks like this, no one has learned to play chess by thinking like this and even trainers and authors of chess books don't think like this." He tested the maxim that a wing attack is best met by a counterattack in the centre against a database sample — 110 games in which a kingside attacker played g4 on move 17 — and concluded that at most 10% of them call for the central counterattack.
Silman's own answer largely concedes the mechanism: his idea of imbalances, he wrote, "was created as a shortcut to pattern recognition. Since most students were not willing to do it the old fashioned way, I decided that imbalances would give them a full diet of patterns." Both of them agree the underlying skill is pattern recognition; they disagree only about whether verbal frameworks deliver it. Which means the frameworks on this page are scaffolding for pattern acquisition — use them as a substitute for seeing hundreds of positions and they will not work.
The retrieval way to read a strategy book
One book read this way beats six read normally, and it takes about four times as long per game. That is the point. If your reading rate does not drop, you are not doing it.
What to do at each rung
| Rung | Strategy work |
|---|---|
| Under 1150 | None. Genuinely none. Do Rung 1 and 2. The only structural idea you need is "open file, put a rook on it." |
| 1150–1449 | The move-fifteen drill on your own games. Learn the plan of the structure your opening produces, and no others. |
| 1450–1749 | The five-step procedure once a game, in writing. Trade discipline: name the better piece before every exchange. One annotated-games collection, read the retrieval way. |
| 1750+ | Structures in depth — a book like Flores Rios, worked through the exercises rather than read. Model games in your own structures. This is where strategic knowledge starts converting reliably. |
The whole page in one paragraph
Read the plan off the pawn structure rather than inventing it. Name the structure, state what it tells both sides, find a weakness that can actually be occupied, improve your worst piece, ask what they want, then check the move for safety anyway. Trade pieces when ahead and pawns when behind, trade when cramped, and always trade your bad piece for their good one. Take space only when you can name what will use it, and break a rule only with a calculated line in hand. Strategy will not save a game you lose to a hanging bishop — and once you have stopped losing those, it is most of what is left.
Figures on this page with their weaknesses: Kaufman's piece values (~300,000 master games at FIDE 2300+, drawn from a larger set of about 925,000, published on chess.com with methodology stated); Solon's space analysis (100,000 Lichess games, single unreplicated blog analysis, Stockfish's space term, chart rather than published numbers); Hendriks's 110-game sample and the Silman reply (both reached us through one review, we have not seen the book or the original column); Chessable's candidate-moves null result (207 players, ten puzzles). Every move sequence above was replayed through a chess engine for legality and the resulting pawn skeletons checked square by square. The claim most likely to need revising is the one this page opens with — that strategy is close to irrelevant below about 1150 — because nobody has measured it, including us.