Battleship strategy: search like a computer

Battleship is a search problem

You cannot influence where your opponent's ships are. Everything you control is where you look and where you hide — and a good searcher beats a bad one by about fifty shots a game.

The game has exactly two modes, and strong play means knowing which one you are in:

1. Hunt — no unresolved hits on the board. Fire where ships are most likely to be.
2. Target — you have a hit that is not part of a sunk ship. Finish that ship before anything else.

Level 1 — Basic: hunt and target properly

1. Use parity when hunting

Every ship is at least two squares long, so it must cover at least one square of any checkerboard pattern. Fire only at squares of one colour while hunting and you halve the search space at no cost. Once only ships of length three or more remain, you can space your shots even further apart — every third square.

2. Switch to targeting the moment you hit

Two rules cover it:

• One hit — fire at the four orthogonal neighbours.
• Two hits in a line — fire at the ends of that line. The ship cannot bend.

3. Finish before you hunt again

Leaving a wounded ship to hunt elsewhere wastes shots: you will have to come back, and in the meantime your unresolved hit tells you nothing.

4. Do not bunch your fleet

Ships packed in one region die together, because any hit in that area leads straight to the next ship. Spread them.

5. Avoid the middle for your biggest ship

Central squares are covered by more possible placements than edge squares, so an opponent searching intelligently looks there first. The carrier is the easiest ship to find — give it the least likely place.

Level 2 — Intermediate: probability density

The strongest practical method is to ask, for every empty square: how many of the possible remaining ship placements would cover it? Fire at the highest number. This is probability density, and it is what strong Battleship programs do.

What the map tells you

• Centre squares are far more likely to be occupied than corners, because more placements fit through them. Shoot centrally, hide peripherally.
• The map changes after every shot. A miss removes every placement through that square, which can dramatically raise the probability of its neighbours.
• Sinking a ship changes everything, because that length leaves the calculation.

Using it without doing arithmetic

You do not need to compute numbers at the table. Two habits capture most of the gain:

1. Hunt near the centre first, on parity, then work outwards.
2. After misses, notice which regions still have room for the longest remaining ship — if the carrier is unfound and only two stretches of five squares remain unshot, shoot there.

Deployment, the other half

• Use the edges — but not every ship, or that becomes your pattern.
• Avoid touching ships even when the rules allow it: when one is found, the other goes too.
• Split lengths across the board so that a discovered region does not contain half your fleet.
• Vary between games. Human opponents remember.

Level 3 — Advanced: adaptive search and modelling your opponent

Track the possibility space, not the board

Advanced play means maintaining, in your head or on paper, a mental list of where each unfound ship can still be. The moment a ship has only one or two possible placements left, you have a guaranteed hit — and those situations arise far earlier than players expect, usually once two or three ships are sunk.

Exploit the sunk-ship announcement

Standard rules make the defender say which ship sank. That is a large amount of information:

• It tells you the exact length removed from the pool.
• It confirms the orientation of the squares you hit.
• Combined with your misses, it often eliminates whole regions for the remaining ships.

If you are playing the harder variant where sinks are not named, keep hitting: an "extra" hit beyond a ship's length proves two ships are adjacent.

Model the opponent, not the grid

Humans are not random. Real deployments cluster in predictable ways:

• Most players avoid the outermost ring, feeling it is "too obvious" — which makes the second ring in unusually crowded.
• Most players avoid touching ships, even when allowed.
• Most players place at least one ship along an edge.
• Very few players place multiple ships in the same quadrant.

If you have played someone several times, their habits are worth more than any general probability map. And the counter is the same as in Stratego: randomise your own deployment on purpose, including occasionally doing the thing you would normally avoid.

Playing the Salvo variant

Salvo — one shot per surviving ship — rewards a completely different plan. Because your firepower shrinks as you lose ships, speed matters more than accuracy early: spread your salvo across the board to find ships fast, rather than probing one region carefully. Conversely, protecting your own fleet is worth more, so spread your ships and give the small ones the least likely squares.

The mistakes that cost shots

1. Hunting without parity. Half your shots are wasted.
2. Abandoning a wounded ship. You always have to come back.
3. Shooting the sides of a two-hit line. Ships do not bend.
4. Shooting corners early. They are the least likely squares on the board.
5. Clustering your own fleet. One discovery becomes three.
6. Using the same deployment every game against the same opponent.
7. Ignoring the sunk announcement — it is free information that narrows everything.

Common questions