Solver Says
Five Wordle positions a day. Your only job: play the best next guess.
Loading today's positions…
Pick the guess you would play next:
Your recent days
Ratings are stored only in this browser. Taller bars are better days; the dotted line marks 400, the level where most moves match the solver.
What this game trains
Solver Says isolates the one decision that most separates strong Wordle players from lucky ones: the middle-game guess. Each position shows a real opening line and the exact colors it produced, plus how many words in the site's working answer pool still fit. Your move is graded by one measure — the number of candidates it leaves, on average, across every answer still possible. That is the same arithmetic behind the site's solver math: a guess that splits the survivors into many small feedback groups beats one that leaves them in a lump. The recurring lesson is the trap each position contains — the familiar word that feels safe but barely splits the list. Playing five of these a day builds the habit of counting before guessing.
How grading works
Every accepted guess gets a value: the expected number of candidates left after Wordle replies, computed over the position's candidate list with each candidate weighted equally. Your grade is the gap between your guess's value and the best value available in that position — "best" meaning the lowest expected remainder among all 14,855 accepted guesses, which is the solver's measure rather than a claim about perfect play.
| Grade | Gap vs the best move | Points | What it means |
|---|---|---|---|
| Brilliant | ≤ 0.05 | 100 | You matched the solver's top move, or tied it within rounding. |
| Good | ≤ 0.75 | 75 | Less than one extra candidate left on average — a solid, defensible move. |
| Inaccuracy | ≤ 2.5 | 40 | Your guess leaves noticeably more work: up to a few extra candidates on average. |
| Blunder | > 2.5 | 10 | The list barely shrinks — usually the trap word that felt safest. |
A day's rating totals the five rounds, so 500 means five Brilliants. The gaps are measured in candidates, not points of style: a Blunder with 30 candidates left genuinely costs you a turn more often than not.
Positions are built offline from the site's 2,360-word working answer pool and its reconstructed accepted-guess list. Candidate counts describe this site's pool — they are not a claim about what the NYT will choose. Openers in the deal mix the most popular starts with the strongest ones from the site's opener statistics.