Lesson 5 of 8
Memorizing a position: chunking and reconstruction
You do not need a better memory. You need to learn how to CUT a position into meaningful groups.
Practise it now, free
The problem to solve
Holding a whole position (up to 32 pieces) in your head sounds impossible. Yet it is THE central blindfold skill: without a stable image of the position, you cannot calculate or follow a game. Typical task: look at this position for 8 seconds, then rebuild it eyes closed.
The key is not superhuman memory. It is the way you ENCODE the position.
The science: chunking
Two founding experiments, worth knowing as they are:
- de Groot, then Chase and Simon (1973): shown a real game position for five seconds, a master rebuilds nearly all of it. A beginner gets a handful of pieces.
- Chase & Simon (1973): rerun the experiment with pieces placed at random. Now the master does no better than the beginner.
The capital conclusion
The master does NOT have better raw memory. He groups the position into "chunks", packets of 4-5 pieces that mean something (a castled king, a pawn chain). Working memory holds only a few units, three or four for most people; the master turns thirty pieces into a handful of chunks, so it fits. On a random position there are no chunks to form, and he becomes a beginner again.
The takeaway: you do not need a better memory, you need to learn to CUT. That is both liberating and a method.
Technique 1: the pawn skeleton first
Pawns move rarely and structure everything. Encode them FIRST, like the bone structure: pawn chains (b2-c3-d4), doubled, isolated and backward pawns, open and half-open files, majorities on one wing.
Once the skeleton is in place, pieces hang onto it ("the bishop behind the chain", "the rook on the open file"). Far more stable than memorizing 8 pawns one by one.
Technique 2: standard chunks (one word = one group)
Learn a vocabulary of recognizable packets:
- "Kingside castle" = 6 pieces at once: king on g1, rook on f1, pawns f2-g2-h2, often a knight on f3. One word places the whole corner.
- "Fianchetto" = bishop on g2/b2 plus the reshaped f-g-h or a-b-c pawns.
- "e4-d4 pawn center", "French chain e6-d5", "queen + bishop battery", "doubled rooks on a file".
You build your chunk library by playing and studying. The more patterns you know, the less there is to remember.
Technique 3: the glimpse test (the central exercise)
The glimpse is the gym, the game is something else
In a game you never memorize thirty-two pieces again after every move. You hold the last position and apply one move to one chunk: the piece leaves its square, lands on another, and you check what that opened. Rebuilding from scratch is training, updating is playing.
The daily drill that trains everything:
- Look at a real position for 5 to 10 seconds.
- Hide the board.
- Rebuild it on an empty board or in your head.
- Say the threats, weaknesses and plans out loud.
Increase difficulty in steps: first half the board, then the whole; first square occupation only, then the exact identity of each piece.
Golden rule
Always train on REAL positions (from games), never random ones. Chunking only works on meaning.
Technique 4: the memory palace (for long-term storage)
To memorize positions or whole lines over time (openings, famous games): build a known mental route (your home, a commute), place each chunk at a stop as a vivid image (PAO system: Person-Action-Object), and to recall, walk the route and pick the images back up. Memory champions' method. Heavier to set up, reserved for long-term storage, not for in-game calculation.
Advanced variant: some players split the board in two (White's side, Black's side) and encode each half as a short letter string turned into an image. Powerful but it requires a personal system; for mnemonics enthusiasts only.
Technique 5: understand rather than memorize
The ultimate shortcut: when you know WHY a piece is there, you barely have to memorize it. Meaning does the recall. A well-understood position (who attacks what, what the plan is) is retained almost for free, because every piece has a reason to exist. Hence: always verbalize the "why" during the glimpse test.
A fixed scan order (so nothing gets forgotten)
Three counts that catch a broken memory
Before you trust a recall, check three things: the material on each side, the number of pawns on each file, and the color of each bishop. A bishop that changed color, or a ninth pawn, means the position in your head is already wrong, and no amount of care downstream will fix it.
- Pawns of both sides (the skeleton).
- Kings (and castle type).
- Rooks, then bishops (by color), then knights, then queens.
A constant order prevents omissions and becomes an automatism. One retrieval order cited by mnemonists: rooks, dark-squared bishops, light-squared bishops, knights, queen, king.
Try it: the glimpse trainer
Toggle the chunk overlay to see the position cut into colored groups, then run the timed study + reconstruction with a score.
Recommended training progression
- Glimpse test on half a board, occupation only (which squares are occupied).
- Glimpse test on the full board, occupation.
- Glimpse test with piece identity.
- Increase the hidden delay before reconstruction (defer the recall).
- Rebuild + state the threats (bridge to the tactics lesson).
Common mistakes
- Memorizing square by square. The fundamental error: you saturate the 7 memory slots with 7 pieces. Cut into chunks first.
- Training on random positions. Useless: no chunks to form. Always real games.
- Skipping the pawn skeleton. Without the frame, pieces "float" and blur together.
- Not verbalizing the why. Understanding is the best fixative; a mute position fades fast.