Squareword Solver — Finish a 4×4 Word Square

What Squareword Is

Squareword is a daily word puzzle built on a word square: a grid in which the words read across and the words read down are both real words. The board is 4×4, so there are 16 cells, four rows and four columns. Solving it means finding four four-letter words for the rows and four four-letter words for the columns, with every letter shared between one row and one column.

There are no clues and no colour feedback. Each square holds one letter, and that letter has to work in two directions at once. That single rule is the whole puzzle. A crossword lets you answer a clue you know and come back to the rest; a word square does not, because a letter that completes a row can destroy the column running through it. The satisfaction comes from watching a grid lock into place when the two directions agree.

Word squares are old. The Sator square, a five-word Latin square reading SATOR AREPO TENET OPERA ROTAS, survives in the ruins of Pompeii, and English four-by-four squares have been a puzzle-book staple for well over a century. Squareword's contribution is packaging the idea as a small daily grid with a handful of letters already filled in.

The name is sometimes typed as squareward, or split into square word; all three refer to the same 4×4 puzzle. Do not confuse it with a five-letter guessing game. There is no hidden target word in Squareword and no six-guess limit; the letters already on the board are the entire set of clues. You are not guessing a word, you are completing a shape.

The Grid Is 4×4, Not 5×5

The most common mistake is a five-letter habit. Wordle and its many relatives use five columns, so players type five letters into a row and wonder why the last one will not fit. Squareword has four columns. A row holds exactly four letters, and any fifth keystroke belongs to the next row or to nothing at all.

Count the words while you are at it. A 4×4 grid contains eight words: rows 1 to 4 across, and columns 1 to 4 down. Sixteen cells, eight words, every letter counted twice. A five-by-five grid would hold ten words, and the crossing pattern would be looser because each cell would still belong to two words but each word would carry more slack. Four letters is tight enough that almost every letter you place removes a large share of the remaining options.

That tightness is why the puzzle works with only six givens. In a 4×4 grid, six letters is more than a third of the board, and almost every one of those letters narrows two words at once. In a five-by-five grid the same six letters would leave far more freedom.

Why Every Puzzle Is a Mirror

The 2,000 puzzles in the site's bank share a property that changes how you solve them: every one is symmetric. Every entry in the bank passes that mirror test, with no exceptions. Symmetry here means the grid is a mirror across its main diagonal, so the cell at row r, column c always holds the same letter as the cell at row c, column r.

The practical consequence is that row 1 is the same word as column 1, row 2 is the same word as column 2, and so on. A row you complete is not only a row: it is also a column you have completed. Four correct letters across the top place the first letter of all four down-words at the same time, because those down-words begin with the four letters you just wrote.

Symmetry also reduces the amount of independent information on the board. Sixteen cells look like sixteen unknowns, but a mirror grid has only ten independent cells: the four cells on the diagonal, plus one cell from each of the six pairs that sit either side of it. The other six cells are copies. When a puzzle shows six givens and none of them are mirror partners, those six clues can pin up to twelve of the sixteen cells.

One consequence surprises people. The four diagonal cells are the least informative givens in a symmetric puzzle. A letter at row 3, column 3 belongs to row 3 and to column 3, but those are the same word, so the letter constrains one word rather than two. A letter off the diagonal belongs to two different words and is copied to its mirror cell, so it constrains two words at once. If you have a choice about which given to build around, build around an off-diagonal one.

A general word square does not have to be symmetric. The bank's squares are, and the solver offers symmetry as an option rather than assuming it, because you may be checking a grid from somewhere else.

How to Read the Givens

A Squareword board arrives with between six and ten letters already placed. The distribution across the 2,000-puzzle bank is not flat, and knowing it tells you what kind of puzzle you are looking at:

Six givens is not a hard mode bolted on at the end; it is the normal case, six puzzles in ten. If a board looks almost empty, that is the design working as intended. Fewer givens means fewer constraints to chain together, and that alone makes the solve longer.

Read the givens by line, not by cell. Write down how many letters each row and each column already has. A row with three givens is a short candidate list. A column with one given is a long one. The line with the most letters is where the puzzle will open, and it is usually cheaper to spend your first minute there than to stare at the whole grid.

Also read the givens for position. A letter in the first cell of a row limits the word to those that begin with it, which is a much smaller set than words that merely contain it. A letter in the last cell limits the ending, and English endings are lopsided: in the game's 1,000-word pool, 179 words end in E and 131 end in S, while only five end in C. Position matters as much as the letter itself.

A Worked Example

Here is the pattern the sample button loads. Dots are blanks:

H  .  .  .
.  R  .  .
.  .  A  D
.  .  D  E

Six givens. Four of them sit on the diagonal (H, R, A and E), and the two Ds are a mirror pair, so under symmetry they carry one letter of information rather than two. Rows 3 and 4 end in AD and DE. Columns 3 and 4 therefore also end in AD and DE, because row 3 is column 3 and row 4 is column 4.

Take the first row as HARM, one of the pool's H-words. Then column 1 is HARM, so row 2 starts with A and row 3 starts with R. Column 3 starts with R and ends with AD, so it is R?AD: READ or ROAD. Row 2 already has R in its second cell, so it is AR??. If column 3 were ROAD, row 2's third letter would be O and no everyday word fits AR O. If column 3 is READ, row 2 becomes AREA. Column 4 then starts with M and ends in DE, which gives MADE, and the grid closes:

H A R M
A R E A
R E A D
M A D E

Check it in both directions. Across: HARM, AREA, READ, MADE. Down: HARM, AREA, READ, MADE. The grid is its own transpose, which is what symmetry means. In the everyday pool this pattern has exactly one completion, because every other H-word breaks at least one column; in the wider dictionary it has many more, most of them words nobody would use in conversation. That gap between one answer and dozens is the single best argument for solving with the game's own pool switched on.

How the Solver Works

The solver reads the sixteen boxes, treats empty boxes as wildcards, and searches for grids in which all four rows and all four columns are words from the selected list. It does not guess an answer and it does not know today's puzzle. It works only from the letters you type.

Three ideas keep the search fast. First, every word is indexed by position and letter, so a single given narrows a row to a short list without scanning the dictionary. Second, the rows are filled from the top down, and after each row the solver checks whether the partial columns can still become words. A column that has become an impossible prefix is abandoned immediately, which is where most of the savings come from. Third, when a row is finished, the column letters it creates are tested against the same index, so a bad crossing is rejected before the next row is tried.

The search is depth-first with backtracking, and it stops at the first 24 grids it finds. It also carries two safety limits: about 1.2 seconds of wall time and 1.5 million row placements. If either limit is reached, the page says so instead of pretending the list is complete. A six-given puzzle over the pool usually finishes in a few milliseconds, which is why the result appears the moment you press the button.

Every grid that comes back is re-checked independently: the page rebuilds the four column strings from the returned rows and tests all eight words against the same list. If a check failed, the grid would be flagged rather than shown as a valid answer. The verification is not decoration; it is the difference between a solver and a page that prints plausible-looking letters.

The pool is the frequency-ordered list the game uses to generate puzzles, loaded from the same data file as the puzzle bank; the puzzles themselves are never consulted. If the everyday pool cannot complete your letters, the solver widens to the full dictionary and tells you that it did. If the letters are contradictory, it stops before searching and names the row or column that cannot be a word.

What the Solver Can and Cannot Do

It can complete a 4×4 grid from any consistent set of letters, including a completely empty board. It can tell you when a row or a column has no possible word. It can show several valid grids when your letters allow more than one. It can restrict itself to symmetric squares, which is how the official puzzles are built. And it can search a much wider dictionary when the everyday pool comes up short.

It cannot look up the answer to today's puzzle. It has no connection to the daily board and no list of dates. The only thing it knows is the letters you type, which means a typo in your givens produces a confident, correct answer to the wrong question.

It cannot promise that its first result is the intended one. The everyday pool is the game's own 1,000-word generation pool, so a grid built from it is very likely to be the answer the puzzle expects. The solver still never opens the bank's stored puzzles or compares your board to a stored solution; it uses only the word pool. When several pool grids fit, all of them are legitimate word squares, and the game's uniqueness check is what separates them.

The search list is a general four-letter dictionary, or the 1,000-word creation pool when you leave the default on. Neither is the 2,000-puzzle bank, so a result can be a perfect word square and still not be the grid the game generated for today.

It cannot make the full dictionary behave like the game. That list carries 5,454 four-letter entries, including dialect forms, abbreviations and words that only appear in word-game dictionaries. A grid built from it can be valid English-by-dictionary and still be rejected by the game's checker, which is why the everyday pool is the default.

Finally, it cannot search forever. The 24-grid cap and the time limit are deliberate. A loose set of two or three givens can admit thousands of squares, and printing all of them would be useless as well as slow. The page reports when it stopped early so you can add a letter and narrow the list.

Solving by Hand, in Order

The method below works on any 4×4 word square, symmetric or not. It uses the same constraints the solver enforces, without the brute force.

1. Count the givens on every line. Four row counts and four column counts. The highest count is your entry point, because a line with three letters has a short candidate list and a line with one letter has a long one.
2. Write the candidates for the tightest line. Use the pattern with blanks: H _ _ M suggests HARM or HOLM; HARM is the everyday choice and HOLM is rare. Keep the list short and prefer words you would actually use.
3. Cross the candidates against a second line. Each candidate for row 1 fixes the first letter of all four columns. Reject any candidate that gives a column a starting letter with no plausible word behind it.
4. Work the shortest list next. After the first row, one column will usually have two or three letters and a much shorter list than the others. Follow the constraint, not the mood.
5. Use the mirror as a free check. If the puzzle is symmetric and you place a letter at row 2, column 4, the same letter must appear at row 4, column 2. A mismatch means the row you just chose is wrong, and you learn it without spending a guess.
6. Verify all eight words at the end. Four across, four down. A grid that reads beautifully across and produces gibberish down is not solved.

The order matters because each step shrinks the next one. Solving a word square by staring at the blank cells and waiting for inspiration is much slower than solving it as a sequence of small candidate lists. The grid rewards bookkeeping.

Strategy Without a Solver

Start with the letter that begins the most lines. Under symmetry, the first row decides the first letter of every row and every column, so it is the single most valuable line on the board. If the top-left cell is given, your first row is already narrowed to words beginning with that letter; if it is blank, you get to choose, and you should choose a word whose letters all make good openings.

Letter frequencies in the game's pool are lopsided, and they point the same way for both ends of a word. These counts are measured from the 1,000-word pool:

Two lessons follow. E is the safest ending in the pool by a wide margin, and S is both a common start and a common end, which makes it the most flexible letter on the board. Q, Z and U are dangerous in either position: one word in the pool starts with Q, two start with Z, and seven start with U. If a candidate row would put Q at the start of a column, you need to know the word that rescues it before you commit.

B and C deserve a warning of their own. They are strong openings (78 and 68 words) but terrible endings (seven and five). A row ending in B or C forces the next row down to start with that letter, and the supply is thin. Prefer to place B and C at the start of a line rather than the end.

The longest crossing word is the one to attack first. By that I mean the line that shares the most cells with lines that are already partly filled, not the line with the most blanks. A row of four blanks that crosses three nearly complete columns is easy; a row of four blanks that crosses four empty columns tells you nothing. Count the intersections, not the letters.

When you are stuck between two candidates, test the one that creates more common column starts. If one candidate makes column 2 begin with S and the other makes it begin with U, the S candidate has a hundred words behind it and the U candidate has seven. You are not choosing the prettier word; you are choosing the branch with more room.

Finally, use the diagonal deliberately. In a symmetric puzzle it carries no information you do not already have from the rows, so it is a checking line rather than a clue. Read it last and use it to confirm what the rows told you. In an asymmetric square the diagonal does real work, because each diagonal cell belongs to two genuinely different words.

Common Mistakes

Thinking it is 5×5. This is the most common error by a distance. Five columns do not exist, a fifth letter in a row has nowhere to go, and any reasoning built on five-letter words is aimed at a different game. Count the boxes before you count the words.

Forgetting the columns. People fill the rows, see four readable words, and submit. Half the puzzle is down. A grid can have four perfect rows and four impossible columns, and that grid is wrong. The columns are not a tiebreaker; they are equal partners.

Treating a cell as belonging to one word. Every one of the sixteen cells sits in a row and a column at the same time. Typing a letter because it finishes a row without asking what it does to the column is how grids break. The solver's whole advantage is that it never forgets this; a human has to remember it on purpose.

Assuming any dictionary word is accepted. The game checks against its own pool of everyday four-letter words. An obscure but genuine word from a large dictionary can be rejected even though it is spelled correctly. If the game refuses a grid that looks valid, this is the usual reason.

Assuming every word square is symmetric. The bank's 2,000 puzzles are, but the definition of a word square does not require it. If you are checking a square from a book or another site, do not assume row 1 equals column 1 until you have checked.

Trusting a nearly complete grid. Three good rows and one doubtful row is not a solved puzzle. Fill the doubtful row, then check all four columns again; changing one letter can fix one column and break another.

Ignoring the mirror check. In a symmetric puzzle, a letter at row 2, column 4 must equal the letter at row 4, column 2. Skipping that check throws away a free test that costs nothing and catches wrong rows early.

Checking a Finished Grid

Before you submit, do the same eight checks the solver does. Read the four rows left to right, then read the four columns top to bottom. Each line must be exactly four letters and each must be a word in the game's pool. Then, if the puzzle is symmetric, confirm that the grid equals its own transpose: row 1 must match column 1, row 2 must match column 2, and so on. A symmetric puzzle that fails the transpose test is wrong even if every line happens to be a word.

One last check is worth the two seconds: read the grid aloud. A square that reads HARM AREA READ MADE sounds like four words. A square that reads HARM AREA REED MADE also sounds like four words, and it is a valid square, but the game's uniqueness check decides which one the puzzle intended. If both are real words in the pool, you have found an ambiguity rather than a mistake, and the only way to resolve it is the game's own answer. The solver shows you the alternatives instead of hiding them.

Related Squareword and Word Tools

If you want to play first and solve later, open the Squareword game. The Squareword rules and strategy guide explains the modes and scoring, and today's Squareword hints covers the daily board. For other four-letter work, the word finder searches by pattern and length.