Guess order: adapting to feedback

You cannot choose the hidden answer, but you can choose a guess that asks a useful question about it. Compare the candidates your feedback could distinguish, then decide whether to gather information or attempt a solution. Both the answer and your decisions affect the result; a method does not remove uncertainty.

Luck versus information

Two things decide how a game goes. One is luck: the answer is different every day, and some days your opener happens to land more clues than others. The other is information — how fast you shrink the list of words that could still be the answer. A short run of scores does not isolate those effects. A useful comparison examines the possible feedback for the same candidate set, instead of attributing a good or bad week entirely to strategy.

Every guess is really a question

It helps to think of a guess less as “a try at the answer” and more as a question about the word. Choose the objective before comparing guesses: a smaller average remaining group, a smaller worst-case group, or a greater probability of solving within the turns left. These are different objectives. A word can improve the average while leaving a larger worst branch; neither statistic alone establishes the probability of winning, which also depends on later choices. Common, distinct letters can provide broad coverage at the start. Later, a repeated letter or a change of position may answer the unresolved question more directly. Compare the size of the groups a guess creates, including the group you would least like to receive.

Why the order matters

Early information gives you more opportunities to respond. A different-letter follow-up can help after an empty row, but useful misplaced letters may call for a position test instead. The value of order comes from adapting: read the first result before choosing the next word. A fixed pair reveals the same combined constraints in either order if both guesses are submitted; the advantage appears when the first row changes your plan or solves the game before the next guess is needed.

The luck you cannot control — and the part you can

You cannot choose the answer, and you cannot force a lucky opener. What you can control is your information rate: how much each guess is set up to teach you. A candidate calculation can compare that information without making claims about player averages. It does not guarantee a solve, and a low-information word can still happen to be the answer.

In practice: open with a reliable first word (see the starting-word comparison), respond to its feedback with a considered second guess, then feed what you know into the Word Solver to see the words that still fit. Want to practise the pattern before the shared daily word? Unlimited mode lets you run it as many times as you like, and the tips page collects the rest of the tactics.

Why more feedback patterns can still leave larger groups

These comparisons use the 1,402-answer pool checked on October 17, 2026, giving each answer equal weight. Pattern counts describe distinct feedback rows; average remaining counts describe the size of the compatible group, including the solved answer itself.

TRACE: 141 patterns; average remaining 38.3.
STARE: 124 patterns; average remaining 36.9.
CRATE: 136 patterns; average remaining 37.8.

TRACE has more pattern types than STARE, yet a larger average remaining group. More types can coexist with large groups reached by many answers. CRATE uses the same letters as TRACE, but rearranging them changes exact matches and the partition of answers. A letter-frequency total cannot capture that positional difference. This is a comparison of selected openers, not proof that one arrangement wins on every board.

“Average remaining” gives each answer equal weight, looks up the size of its feedback group, and averages those sizes. Large groups receive more weight because more answers lead into them. The solved answer still counts as a compatible candidate. This is a measure of remaining uncertainty after a move, not a full strategy simulation and not a measured average solve length. The opener comparison also shows largest-group sizes, which answer a different question about an unfavourable result.

Compare feedback groups before counting new letters

A letter absent from every remaining candidate returns the same miss for all of them. “New to the keyboard” is therefore not the same as “informative about this shortlist”. Check the actual rows the guess would produce before using novelty as a tie-breaker.

Position can distinguish candidates even when the letter set stays the same. If possible answers place a known letter in different columns, an exact tile may separate them where an ordinary presence test cannot. Repeated letters add another question: how many copies fit? This is why a broad opening habit should give way to a board-specific decision. Once you have clues, your original dictionary-wide frequency ranking is no longer the whole problem you are trying to solve.

Information needs a turn in which to use it

A free probe can be valuable before the endgame because you can act on what it reveals. On the final turn, a guess outside your possible-answer list cannot win. With more time available, compare the branches: which result would isolate a solution, and which would still leave a group too large to try individually? A low average remaining count can hide an awkward branch. A smaller worst group can be useful even when the average is slightly higher, depending on your aim for the game.

Use the solver as a source of accepted words fitting the fields you entered, then verify additional restrictions against the board. It does not choose an optimal decision tree for you. The example below isolates an ordering decision so you can check the feedback yourself in practice.

Change the second word when the first row answers your question

Use STARE, STALE and STAGE as a small demonstration set. All belong to the answer pool. Suppose you prepared the pair STARE then ROATE. If the answer is STARE, the first guess solves it and the second is unnecessary. Against either STALE or STAGE, STARE returns ✓ ✓ ✓ − ✓, leaving those alternatives indistinguishable within this set.

ROATE then returns − − ✓ ↔ ✓ against both remaining words. It tests no distinction between L and G, so completing the prepared pair would leave the same uncertainty. Playing STALE instead solves STALE or returns ✓ ✓ ✓ − ✓ against STAGE. With this complete two-word shortlist and a turn left afterwards, that separates the alternatives. With additional candidates, you must check their rows too before drawing that conclusion.

Reversing the fixed pair does not rescue its final distinction: ROATE produces ↔ − ✓ ↔ ✓ against STARE, and − − ✓ ↔ ✓ against both STALE and STAGE. Submitting STARE next still leaves the latter pair together. The useful change is choosing a different follow-up after reading feedback, not merely swapping the prepared words. For reviewing a finished game, use the mistakes checklist.

Keep exploring

Play today’s Puzzico or browse all guides & strategy.