Word Square Strategy: Solve the 5x5 Grid in Ten Guesses (2026)

Last updated: October 15, 2026 · Tested on Playnesta's Word Square build (engine, UI and the 120-board set are Playnesta-original, MIT; the 12,972-word list is the MIT-licensed wordle-clone by WebDevSimplified @ b12c2df; the interlocking-grid mechanic is modelled on Squardle by Fubar Games and appears here as a genre label only) · 5x5 grid, 21 letter squares + 4 blocks, six five-letter words · 10 guesses, +1 per solved line, 5 bonus maximum

Method: every number below comes from this site's playable build and was verified this session, not quoted. The headless run passed 50/50 checks on pinned board 42 (rows ABOVE / CABIN / PAUSE, columns BARRA / VIRUS / ENSUE) plus boards 0 and 7: a 5x5 board of 25 squares = 21 letter squares + 4 blocks holding six five-letter words — three rows and three columns whose 3x3 crossings agree by construction and all six of which are inside the accepted word list; a start of 10 guesses with 0 bonus and a header that reads "10 guesses left"; the fixed line order c1, c3, c4, r0, r1, r4 on board 42, with the header naming the active line (Column 2 first); a word outside the list refused with the guess count and the typed buffer both untouched; a four-letter buffer refused as "Not enough letters"; a typed letter that disagrees with a locked green square refused; guessing ARISE against the column BARRA producing orange "1/1" on A (in this column and in the crossing row), yellow "1" on R, white on I and S, and red "1" on E — exactly the rule the game page states; an all-off-board word (GHYLL) marking every square black; a solved line granting exactly +1 guess and handing the turn to the next unsolved line, while an unsolved line stays active; all six lines cleared for a win with the bonus capped at 5; a burned-out budget ending as a loss with no line solved; ?board=42 reloading the identical board and New board stepping to the next one and resetting to 10 guesses; all 120 boards machine-valid (five letters, in the list, no repeats, 21 letters/4 blocks, crossings agree); both production stages 348x261 and 894x670 rendering 25 squares with the keyboard inside the viewport, no horizontal scroll, 44px tap targets and zero console errors; and the physical keyboard, the on-screen keys and the bridge's five events all on the same input path. Controls: type five letters and press Enter, Backspace deletes, the on-screen keyboard works identically, and ↻ New board deals another board. The game page explains how to operate the board; this guide is how to win it.

1. You own one line at a time — and it only changes when you solve it

You never pick the line. The board has six five-letter words — three rows and three columns — woven through 21 letter squares (the other 4 of the 25 squares are blocks), and the header tells you which line is active. On the pinned board the fixed order is c1, c3, c4, r0, r1, r4, so board 42 opens on Column 2. The measured behaviour is stricter than it first looks: a line stays active until you solve it. In the measured run a guess that landed no green square left the same column active, and only a fully green line handed over to the next unsolved one.

This is the single most useful thing to internalise, because it removes the temptation to spray. Beginners read "six words" and try to learn something about all of them at once; the build will not let you. Every guess is spent on the one active line, so the question is never "which line is most informative" — it is "what is the best word for this line". Commit to the line you are given and make each guess count.

2. Every hint is about two lines, not one

The colours look like Wordle's, but they carry twice the information, because a letter square that sits in a row usually also sits in a column. In the measured column (BARRA on board 42), guessing ARISE produced all five colours at once, and each one is a separate read:

A → orange "1/1": the letter is in this line and in the crossing line. Orange is the best square you can get short of green, because it simultaneously narrows two words.
R → yellow "1": the letter is somewhere in this line, but not in the column that crosses this square. Yellow is a line-only constraint.
I and S → white: the letter is nowhere in this line, yet it is somewhere on the board. White is a board-level fact: the letter exists, just not here.
E → red "1": the letter is not in this line at all, but it is in the crossing line. Red is a free constraint on a word you are not even guessing.
Black: the letter is not on the board anywhere. A word made of off-board letters (the measured GHYLL) blacks out all five squares — the biggest information gain a single guess can give.

The practical consequence: red is information you did not pay for. When the active line's answer has no E but the crossing column does, you have just learned something about a word you will meet several lines later. Read red squares as notes about the crossing line, and write those letters down. Conversely, white and black are the two cheapest ways to shrink the letter set for both lines at once — a guess stuffed with letters you suspect are off the board is rarely wasted even when it solves nothing.

The digit on a hint counts arrow pairs — the lower of the letter's occurrences in your guess and in the target line. Orange shows both counts as line/crossing (the measured A printed "1/1"); yellow and red show the single relevant count. A digit of 2 means the letter appears at least twice in that word, which is exactly the shape of clue that separates a word like ALLOT from a word with no repeats.

3. Green squares are the engine of the whole board

Solving a line does three measured things at once: every square in it turns green, you gain exactly one bonus guess, and the turn moves to the next unsolved line. The bonus matters less than the third effect the bonuses produce: a green square in a row is also a green square in the column that crosses it, and vice versa. Six words share 21 squares, so the lines are not independent puzzles — the crossing cells are free letters handed to you by a line you already solved.

That is why the board collapses once the middle lines fall. In the measured win, solving a column handed over to the next unsolved line, and the greens already on the board meant the next word could be completed without re-deriving anything. Plan for it: when a line has one or two letters left, it is usually cheaper to spend a guess finishing it — banked bonus plus crossing greens for later — than to keep fishing in a line that is still wide open.

4. The budget is 10 + 5 = 15 guesses for six words — do the arithmetic early

You start with 10 guesses and every fully solved line returns one, up to 5 bonus — so 15 guesses is the hard ceiling, and the measured win finished with 8 left and all 5 bonuses used. Six lines against a maximum of 15 guesses is roughly 2.5 guesses per line, and because bonuses only arrive from solved lines, an unsolved line is a line that is costing you budget without ever paying it back.

So the loss condition is arithmetic, not knowledge. In the measured loss the budget ran to zero with no line solved at all — ten valid words, none of them an answer. The rule that follows is blunt: if a line has consumed three or four guesses without a single green square, change your approach, not just your word. Try a word built from letters the board has told you nothing about yet, so the line either opens up or you learn the line is not where the board's letters are. Spending five guesses on a cold line is how boards that were winnable in 15 become unwinnable in 10.

5. Word choice: the list is 12,972 words, and the answers skew everyday

Every guess — including the composite of locked and typed letters — must be in the accepted list of 12,972 five-letter words; anything else is refused with the guess count untouched (the measured rejection leaves both the budget and the buffer exactly as they were). The answers themselves are not exotic: across the 120 boards' 720 words, 86% sit in the 2,315-word common list that also feeds the word-guess game on this site, and 16 boards are made entirely of common words. So when you are stuck, the answer is much more likely a plain word than an obscure one — guess the boring word first.

Two habits pay for themselves immediately. First, open on a word whose letters you have no information about; the white and black squares it produces do more work than another vowel-heavy guess on a line you already know something about. Second, remember the buffer is positional: a locked green square occupies its slot, so you retype it as part of the five letters, and the composite has to agree with it — typing around the locked square is the single most common rejection (the build refuses a disagreement before it costs you a guess, but it still costs you the keystrokes and the rhythm).

FAQ

How is this different from the rules on the game page?

The game page explains how the board works: three rows and three columns across 21 letter squares, what each colour means, the arrow-pair digits, the bonus guess, and how ?board=42 pins a board. This guide assumes you know that and plays the layer above it: how to spend the budget, why red squares are free information, and when to abandon a cold line.

Can I choose which row or column to guess?

No. The line is chosen for you and stays active until you solve it — the measured build kept the same column active after a guess that landed no green square, and only moved on when a line went fully green. Pick the word, not the line.

What does the digit on a hint mean?

It counts arrow pairs: the lower of how many times the letter appears in your guess and how many times it appears in the target line. Orange shows two numbers as line/crossing (the measured A on board 42 printed "1/1"); yellow and red show the one that applies. A 2 tells you the letter is repeated in that word.

Why did my guess get refused when I thought it was a word?

Two measured reasons. Either the word is outside the 12,972-word accepted list, or your typed letters disagree with a locked green square. Neither costs a guess — the build checks before spending — but the buffer is positional, so retype the locked letters as part of your five letters.

How many guesses do I really have?

Ten to start, plus one for every fully solved line, capped at five bonus — fifteen at the absolute most. The measured win used all five bonuses and still finished with 8 guesses spare; the measured loss burned all ten with no line solved. Six words against fifteen guesses means about 2.5 guesses per line.

Where can I play it?

Right on this site's word-square page. The engine, the UI and the 120-board set are Playnesta-original (MIT, see LICENSE.txt in the package); the 12,972-word list is reused from the MIT-licensed wordle-clone by WebDevSimplified (commit b12c2df). The interlocking-grid mechanic is modelled on Squardle by Fubar Games — a third-party game name used here only as a genre label; no code or data from it ships here.