Battle Grid

Rules

Each side's 5-ship fleet is placed for you automatically at the start of every game — there's no setup step. Your own board (Your Fleet) shows exactly where your ships are; the other board (Enemy Waters) starts blank, since you don't know where the enemy fleet is hiding.

On your turn, click a cell on Enemy Waters to fire at it. A hit marks that cell and you go again; a miss ends your turn. Sink every cell of a ship to sink the whole ship — sunk ships are revealed on the board. Sink the entire enemy fleet first to win.

New to this one? Medium is a good place to start — once it lands a hit, it follows up on nearby squares about 70% of the time rather than hunting every ship down with Hard's full parity strategy, so there's still real luck in play either way.

In 2-Player mode you're sharing one device — after a turn ends, a "pass the device" screen covers the boards so the next player can't see where the other side's ships are before it's their turn.

Controls

Enemy Waters is labeled like real Battleship: columns a through g left to right, rows 1 through 7 top to bottom.

With a mouse or touchscreen: tap a cell on Enemy Waters to fire at it.

With a keyboard: type a column letter followed by a row number, like "b4," into the box below the boards, then press Enter or Fire to confirm.

Type status (or just the start of it, like "sta") and press Enter or Fire at any time -- even when it isn't your turn -- to have both fleets' status read back to you.

Roots

Battle Grid is this collection's take on the classic hidden-fleet guessing game, which traces back to paper-and-pencil play from well over a century ago before it was ever sold in a box. This version keeps things quick by placing both fleets automatically and using a smaller 7×7 board with a lighter 5-ship fleet, rather than the larger grid and longer fleet of the game it's modeled after.

Under the hood

Easy has no memory at all — every shot is a fresh random guess across the board, whether or not a previous shot already found a ship. Medium remembers: if it's currently sitting on an unsunk hit, there's a 70% chance it fires at one of the untried squares next to it, and a real 30% chance it ignores that hit entirely and fires randomly anyway. Hard plays the textbook strategy. Once it's found a hit, it checks whether two or more hits already line up, and if they do, it fires at the next square continuing that line rather than guessing randomly around it. With no hits to work from at all, it switches to a parity search — firing only on squares where the row and column add up to an even number. Any ship 2 squares or longer (the 3, and both 2s in this fleet) has to span both parities no matter how it's placed, so this alone finds three of the five ships without ever wasting a shot on a square that couldn't be part of one. The two 1-square ships can land on either parity, so once every even-sum square has been tried, it falls back to searching whatever's left.

Firing first turns out to barely matter here: 500 simulated games between two medium AIs came out 52.8% first-to-fire, 47.2% second-to-fire — close enough to call it even.

Where Battle Grid genuinely struggles is the same place every hidden-information game does. Hourglass stays interesting because a solved game can still surprise a player who hasn't learned to watch for its hot cells — the ceiling there is about what a person notices, not what's left to compute. Battle Grid's ceiling is a different kind of wall: once both players are running the same hunt-parity-target pattern Hard already plays, there's nothing left to out-think. The rest of the outcome comes down to where the other fleet happened to be hidden and which random square gets hit first — variance no algorithm, human or otherwise, can search its way around.