Results
Guess distribution
A dash hides a low-count result; a second cell may also be hidden so subtraction cannot reveal it.
Open human-performance data
The Wordle Observatory
A public view of how people solve Wordle, built from games they explicitly choose to contribute after a review. It stores a few game facts—not grids, answers, guess sequences, accounts, IP addresses or cross-game identifiers.
6opt-in games
5puzzles observed
—solve rate
—average turns when solved
Privacy threshold in progress
6 of 10 games collected
Counts can show that the project has begun, but outcomes and averages stay hidden until the overall sample reaches the same minimum used for public subgroups.
Playing conditions
What contributors reported
- Hard Mode share
- —
- Average difficulty opinion
- —
- Average decision efficiency
- —
- Minimum public subgroup
- 10 games
These are self-selected contributions, not a representative estimate of every Wordle player.
Method and privacy
Tiny records, guarded publication
Each contribution contains only the puzzle number, result turn count (or X), optional Hard Mode status, optional opener, optional 1–5 difficulty opinion, and optional decision-efficiency bucket. A random per-game token is hashed solely to make retries idempotent.
No answer, colored grid, later guesses, account, email, browser fingerprint, user-agent or reusable player identifier is stored in the observatory table.
Publication rules
- Contributing is off until a player ticks an explicit consent box.
- Live puzzles remain hidden until they are finished in every timezone.
- Puzzle and opener rows appear only at 10 or more games.
- Each optional average or share needs 10 actual responses of its own.
- Low result cells—and a complementary cell when needed—remain suppressed.
- Decision efficiency is rounded to five-point buckets.
- Public CSV contains aggregates only.
- Records are self-reported and cannot be independently verified; rate limits and per-game retry tokens reduce casual duplication.