The Chain Rule in Dots and Boxes, Explained Simply

There is a single idea in dots and boxes that separates players who lose 30% of the time from players who lose 70% of the time. It is called the chain rule (or long chain rule), and it predicts, with surprising accuracy, who is going to win any given endgame. Most people who have played dots and boxes for years have never heard of it. Once you understand it, the way you play changes completely.

This post is dedicated to the chain rule: what it is, why it works, how to count chains during a real game, and what to do when the rule tells you that you are losing.

The one-sentence version

To position yourself to win a game of dots and boxes, try to make the total number of long chains plus loops equal to the same parity (even or odd) as a specific target that depends on who moved first and the dimensions of the board.

That is the compressed version. The rest of this post is about unpacking what that means, why it is true, and how to use it.

What is a chain?

Before we can talk about the chain rule, we need a precise definition of a chain. In dots and boxes, a chain is a sequence of adjacent boxes where, if one box is opened (meaning its third side is drawn), your opponent can complete that box on their turn, and the act of completing it also opens the next box in the sequence, which they can then complete, and so on.

A long chain is a chain of three or more boxes. A short chain is one or two boxes. A loop is a chain that closes back on itself so that the start and end are the same box, forming a ring.

The reason long chains matter and short chains do not is that long chains enable the double-cross. In a long chain, the player receiving the chain can take all-but-two boxes and then sacrifice the final pair to force the opponent to make the next move elsewhere. In a short chain of one or two boxes, there is nothing to double-cross with — the player just takes them and moves on. So short chains are strategically neutral; long chains are strategically decisive.

What does "parity" mean here?

Parity just means "even or odd." Two numbers have the same parity if they are both even or both odd. They have different parity if one is even and the other is odd.

The chain rule is fundamentally a parity argument. It says that whether the total number of long chains plus loops is even or odd determines, under perfect play, which player benefits from the long chain phase.

Here is the most common statement of the rule for a standard 5×5 box grid (the common American version):

The second player wants the number of long chains + loops to be even. The first player wants it to be odd.

For other board dimensions, the parity target flips. On a 5×6 box grid, the first player wants even and the second wants odd. There is a formula — it depends on the total number of dots on the grid — but memorizing every variant is less important than internalizing that one player wants even, one wants odd, and both players are quietly fighting over chain count parity throughout the entire game.

Why does it work? A rough intuition

The chain rule seems magical when you first encounter it. Why would counting chains predict the winner? The answer comes from the structure of the long chain phase.

Here is the intuition. When the game enters the long chain phase, neither player wants to open a long chain first. Whoever opens it gives the other player a huge score — and, via the double-cross, forces themselves to open the next long chain too. This effect cascades. If player A opens the first long chain, player B double-crosses, forcing player A to open the second, and the third, and so on. Player B ends up taking most of every long chain.

The question, then, is: which player is going to be forced to open the first long chain? That depends on how many non-chain moves (neutral moves, short chain moves) are left when the long chain phase begins, and on whose turn it is.

Those numbers are controlled by the parity of the total board and the number of long chains that form. When you work through the arithmetic, you get: if the long-chain count has the "right" parity for your position, your opponent is forced to open first. If it has the wrong parity, you are forced to open first.

The full proof is more elaborate and uses the tools of combinatorial game theory, but the practical picture is always the same: one parity favors the first mover, the other favors the second, and the board dimensions determine which is which.

The formula, for people who like formulas

For a board with $m \times n$ boxes (so $(m+1) \times (n+1)$ dots), the first player wants the number of long chains plus loops to be the same parity as the total number of dots. Since a grid has $(m+1)(n+1)$ dots, you can compute this quickly:

• 5×5 box grid: 6×6 = 36 dots. Even. First player wants long chains + loops to be even. Second player wants odd. Wait — that contradicts what I said above. Let me restate.

Actually, the rule is stated in several different ways across different sources, and the conventions differ depending on who is moving first. The cleanest statement is:

For the player who moves first to win, the number of long chains plus loops should be the same parity as the number of dots.

For 5×5 boxes (36 dots, even), the first player wants an even count. For 5×6 boxes (42 dots, even), the first player wants an even count. For 4×5 boxes (30 dots, even), the first player wants an even count. For 3×4 boxes (20 dots, even), the first player wants even.

Most competitive dots and boxes boards have an even number of dots, so in practice: the first player usually wants an even count of long chains + loops, and the second player wants odd. But before you apply the rule, count the dots on your specific board.

How to count chains during a real game

Counting chains during play is an acquired skill. It is not hard — it is just unfamiliar — and with practice it becomes automatic. Here is the technique:

1. Scan the board for filled-in regions. A "region" is a connected group of boxes where the grid lines have started to form boundaries.
2. For each region, estimate its eventual chain structure. Is it going to become one long chain? Two separate short chains? A loop?
3. Add up expected long chains plus loops across the whole board. Compare against your target parity.
4. If you are on the wrong parity, find a move that will change it.

Step 4 is where the strategy lives. To change the chain count by 1, you can:

• Split a region. Drawing a line that cuts a developing chain into two smaller chains typically adds one to the count (if both halves remain long) or subtracts one (if one half drops below three boxes).
• Merge two chains. If two short chains are adjacent, a move that connects them into a single long chain changes the count.
• Break a potential loop. A move that prevents a loop from closing may convert it into a long chain, changing counts.

Every middle-game move in competitive dots and boxes is, at some level, a vote for or against a particular chain count parity. The stronger player usually recognizes this one or two moves before their opponent does.

What to do when the chain rule says you are losing

Sometimes you count the chains, check the parity, and realize you are on the wrong side. Does that mean you lose?

No. It means you have to play to change it.

The chain rule describes what happens under roughly symmetric play. If your opponent plays reactively rather than proactively — if they are just filling in safe moves without counting — you have the opportunity to manipulate the chain structure without them noticing. Every time you merge or split a region, you are trying to flip parity back to your favor.

Against a stronger opponent who is also counting, parity battles are genuinely hard to win. But against most players at most skill levels, you can exploit the fact that they are not even aware parity exists. Most recreational players will happily let you reshape chain counts because they do not see why the reshaping matters.

A simple example: suppose you are on the wrong parity, with two developing chains of length 2 and length 3 in a central region. You can play a move that merges them into a single long chain of 5 boxes — changing the long-chain count from 1 to 1 (the chain of 3 was already long; the chain of 2 was not), but also changing the structure so that the next move in that region may split it differently. Meanwhile you preserve the ability to threaten a further split later. A weaker opponent will not see what you are doing; a stronger one will block you.

Short chains do not count for the chain rule

A common beginner mistake is to count every chain, including short chains of 1 or 2 boxes. The chain rule only counts long chains (3+) and loops. Short chains are strategically insignificant for parity purposes because they do not allow double-crosses.

That does not mean short chains do not matter at all. They are still worth points, and in close endgames a short chain taken or denied can flip the score. But for parity thinking, ignore them.

Loops: weirder than chains

Loops count for the chain rule the same way long chains do — they add 1 to the combined count. But loops behave differently in two ways:

1. A loop's double-cross costs four boxes, not two. When you receive a loop, you cannot just sacrifice the final two — you have to sacrifice the final four, because closing a loop with a single line completes two pairs of boxes simultaneously.
2. Loops are usually worse to open than equivalent-length chains. Because the opponent's double-cross costs you four, you would rather open a chain of 5 than a loop of 5.

The practical consequence: when counting for the chain rule, treat each loop the same as a long chain (+1 to the count), but when choosing which region to open, prefer opening a long chain over a loop.

A real-game example

Imagine a 5×5 box game. Halfway through, the board has:

• One region on the left that looks like it will form a long chain of 4.
• One region in the middle that could be split into two chains of 2 and 3 (so one long, one short) or kept as a single long chain of 5.
• One region on the right that is shaping up to be a loop of 4.

Current long chain count: 1 (left) + 1 (middle) + 0 (right, but it will become a loop counted as 1) = you project 3 long chains + loops if nothing changes.

If you are on the parity where you want even, you need to either merge two regions into one (reducing the count from 3 to 2) or split one into two pieces where one piece is short (also reducing the count, if the split creates a short chain). You would look for a move that achieves that structural change. Meanwhile, your opponent — if they are counting — will be playing to preserve the count at 3, because 3 is the parity they want.

Every move in that middle phase is fought over this structural question. It is invisible to someone who does not know to look for it, and it is the whole game to someone who does.

Why this matters beyond dots and boxes

The chain rule is specific to classic dots and boxes, but the habit of parity thinking transfers to many grid-capture games. In Dot Clash, for example — which uses a different capture mechanic based on enclosing opponent dots rather than drawing lines — there is no literal chain rule, but there is a related structural question: who is going to be forced to commit to a losing territory first? That question is answered by counting available moves and tracking enclosure parity, much like chain counting in classic dots and boxes.

Once you learn to think in terms of "who will be forced to move next in a region where every move is bad," you start to see games that used to look random as having hidden structure. The chain rule is just the most famous example of this structure, in the game that has been most thoroughly studied.

The three things you should take away

1. Count long chains plus loops throughout the game, and know what parity you want.
2. The parity target depends on dot count — for most standard boards, first player wants the same parity as the dot count.
3. Middle-game moves are parity votes. When you are tempted to play a safe move randomly, check whether that move nudges chain count toward or away from your target.

If you take nothing else away from this post, take this: dots and boxes is not a game of collecting boxes. It is a game of controlling chain parity. The collecting is just bookkeeping. The strategy is the parity fight.

Where to go next

Once you are comfortable with the chain rule, the next concept worth learning is the double-cross — the technique that converts a parity advantage into actual points. There is no point winning the parity battle if you do not know what to do when your opponent opens the first long chain. That is covered in more depth elsewhere on this blog, but the quick version is: when receiving a long chain, take all but the last two boxes and hand the final pair back by drawing a line that completes two boxes at once.

After the double-cross, the next topic is opening theory — how early moves constrain chain structure — and then the endgame, where loops and chains interact in complex ways. But the chain rule is the foundation. Start there, spend 20 or 30 games consciously counting, and you will start winning games you used to lose.