The Endgame Phase: Loops, Chains, and the Final Boundary

Games of dots and boxes, Dot Clash, and most grid-capture games are decided in the final ten to twenty moves. The opening sets up structure. The middlegame controls parity. But the endgame — the phase where almost all captures happen — is where the arithmetic becomes concrete and the winner is decided.

This post is about endgame technique: how to recognize you are in the endgame, how to evaluate remaining regions, how to choose the order in which to resolve them, and how the specific shapes of loops and long chains interact in the final moves.

When does the endgame begin?

There is no clean bell that rings. The endgame begins when every move on the board is either:

• Capturing something (completing a box or enclosing opponent pieces), or
• Opening a region for your opponent (giving them a chain or loop they can take most of).

When there are no more neutral moves — no more moves that leave the board's structure roughly unchanged — the endgame has started. On a typical 5×5 dots and boxes grid, this happens around move 25 (of ~60 total). On a 25×25 Dot Clash board, it happens around move 80 (of several hundred possible).

Recognizing the transition is itself a skill. Beginners sometimes continue to play "safe" moves into the endgame, but there are no more safe moves — every move now gives something up. The right question stops being "which move is safe?" and starts being "which region do I want to be the one to open?"

The three things to count

Once the endgame begins, you should be counting three things on every move:

1. Total captures remaining

How many boxes (or opponent dots) are still capturable on the board? This tells you the maximum score swing possible from here to the end. If only 8 captures are left and you are up by 10, no endgame sequence can cost you the game.

2. Number of long chains and loops

This is the classic chain rule count (in dots and boxes) or the region count (in Dot Clash). It tells you whose parity is favored. If the count is wrong for you, you need to either change it or accept that you will be forced to open first.

3. Available moves per region

This is the subtlest count. How many moves can be made in each remaining region before someone has to commit to opening a chain? This determines who will be forced to open first.

Counting all three at once is hard. It becomes easier with practice. On your first serious endgames, just count one — total captures — and add the others as you get comfortable.

Opening order

In a typical endgame, there are 2–5 distinct regions left. The order in which they get opened determines the outcome. There is a general principle:

Force your opponent to open the biggest region first.

Sounds obvious. It is actually subtle, because "biggest" is not always the same as "most boxes." A 4-box region with a double-cross opportunity is worth more than a 5-box region where no double-cross is possible (because the 4-box region hands back 2 boxes, so the opener receives net 2, while the 5-box straight region hands back 0 and the opener loses 5).

The rule of thumb, refined:

• First, force them to open any region with a double-cross opportunity (i.e., any long chain of 3+).
• Second, among double-crossable regions, force them into the largest one since you gain more (you take N-2 boxes out of a chain of N).
• Save loops for last or avoid forcing them into loops because the opener of a loop loses less (the receiver gives back 4 instead of 2).

In the simplest case, if the opponent has to open the last long chain on the board, you cannot double-cross (there is nothing after), so you just take everything. This is why the last chain is different from every other chain.

The last chain

Every endgame has a "last chain" — the region that, after you resolve it, ends the game. The last chain is special because:

• No double-cross. There is no reason to sacrifice the last two boxes because there is nothing for the opponent to open next.
• Taking it in full is the right move. Whoever takes the last chain wins its full value.

The strategic consequence: the last chain is worth more than any other chain of the same size. A 4-box chain during the middle of the endgame is only worth 2 boxes to you (because you give 2 back on the double-cross). A 4-box chain that happens to be last is worth all 4.

This has a subtle but important implication: the player who receives the last chain benefits more than other chain receivers. In practice, this means the chain rule counting gets modified in the final stages. Instead of counting only long chains and loops, count whether the last chain favors one player or the other.

Loops: the weird case

Loops are long chains that close back on themselves. They differ from straight chains in two important ways during the endgame:

Loops cost 4 to double-cross, not 2

When you receive a loop and want to double-cross, you sacrifice the final four boxes instead of two. This is because a loop's closing move completes two pairs of boxes simultaneously, so a single move hands back 4 boxes.

The economic consequence is big. A chain of 6 boxes gives you 4 if you double-cross (take 4, give 2). A loop of 6 boxes gives you only 2 if you double-cross (take 2, give 4). Loops are much worse to receive.

Loops are worse to open

If you are going to be forced to open something, you would rather open a chain than a loop of the same length, because the receiver of the loop benefits less. From the other side: you would rather force your opponent to open a chain than a loop.

Rule of thumb for loops:

• Double-cross loops of 6+ (trade 4 for tempo).
• Take loops of 4 in full (the double-cross cost of 4 equals the total loop value — no gain).
• Take loops of 5 in full if they are the last loop on the board, double-cross if there are regions left to force.

Sacrifice counting

"Sacrifice counting" is the endgame skill of calculating the net value of each possible opening order. It is the concrete arithmetic that underlies all the heuristics above.

Here is the general form. For each remaining region, compute:

• Region size (how many captures it contains).
• Double-cross cost (2 for a chain, 4 for a loop, 0 for the last region).
• Who opens it (whoever is forced to move into it).
• Who receives it (the other player, who then takes region_size - double_cross_cost).

Then walk through the endgame move by move, figuring out who opens each region in what order. The player with higher total captures wins.

This sounds tedious, and it is, but it is also tractable. With 3 remaining regions of sizes 4, 5, and 6, there are only 6 possible orderings to evaluate, and usually 4 of them are eliminated by the forcing structure (you cannot freely choose who opens which — it is determined by whose move it is when a region becomes the only remaining option).

With practice, you do this counting without explicit enumeration. You look at the board and say "okay, there are three regions, I can force them to open the 6-chain first by playing this move here, I get 4 boxes from the 6-chain with a double-cross, then I force them into the 5-chain, I get 3 from that, then they force me into the 4-chain which I take in full for 4 — total 11 vs their 2+2 = 4." That is the calculation.

Common endgame mistakes

Even among players who know the theory, specific mistakes show up again and again.

Mistake 1: Taking the last-but-one chain in full

If the board has two chains remaining and you are receiving the first one, you must double-cross it — otherwise you will be forced to open the last chain yourself, and the opponent takes it in full. A beginner instinct is to take all boxes from the first chain, not realizing they just handed the game over.

Mistake 2: Double-crossing the last chain

The mirror mistake. If you are receiving the last chain on the board and you double-cross it, you gave up 2 boxes for no benefit, because there is no next region for the opponent to be forced into.

Mistake 3: Miscounting region size

On complex boards, especially with Dot Clash, it is easy to miscount how many captures a region contains. A region you thought was 5 might actually become 7 when you look more carefully at enclosure lines. Miscount the wrong direction and your entire endgame plan is wrong.

Mistake 4: Failing to see a merge

Sometimes two regions that look separate can actually be merged into one big region by a specific move, which changes the capture count and parity. Missing a potential merge is a common mistake because beginners see the board as fixed rather than as something they can reshape.

Mistake 5: Emotional capture-taking

At the end of a long game, the instinct to take any available capture is strong. You worked hard, you see a free 3 boxes, and you take them. Sometimes this is correct. Sometimes it loses you the game because the free 3 boxes were actually a sacrifice setup, and taking them forces you to open a 10-chain. Pause. Count. Then capture.

Endgame in Dot Clash specifically

Dot Clash endgames share the structural features of dots and boxes endgames but with different mechanics. Instead of chains, you have enclosures — regions where your dots are about to close around opponent dots. Instead of the chain rule, you have region-count parity (which player will be forced to place the opening dot of the next big capture).

The double-cross analog in Dot Clash is the partial enclosure — leaving a deliberate gap in your boundary to let a small group of opponent dots escape, so that you force them to reinvest elsewhere. It is less geometrically crisp than a dots-and-boxes double-cross but it is the same idea: give up a little to force tempo.

The endgame skills from dots and boxes transfer to Dot Clash with some adjustment. If you are strong in one, you can get reasonably strong in the other in 20 games of conscious practice.

The three-rule endgame summary

If you want the compressed version of everything above:

1. Count remaining captures, regions, and forcing moves on every turn.
2. Force your opponent to open the largest double-crossable region first; save the last region for yourself to take in full.
3. Double-cross every long chain except the last, and double-cross loops only if they are 6+ boxes or you need the tempo badly.

Those three rules will win you most endgames. Everything else is calibration — knowing when a region's size makes it worth a special exception, when to take a loop in full rather than double-crossing, when to merge regions. Calibration comes from games played, not rules memorized.

Play 30 conscious endgames, counting out loud if you have to. By the end, the arithmetic becomes intuitive, and you will find that games you used to lose narrowly now go your way in the final moves. The endgame is the phase that rewards patient learning more than any other, and it is where the best players feel most at home.