On this page · 7 sections
Equation Dataset#
NerdleBuddy generates mathematically valid eight-character equations that follow the modelled Classic Nerdle format: one equals sign, valid arithmetic, whole-number values, and standard order of operations. The generated file records whether each equation is eligible as an answer or only as a guess.
| Dataset | Count | How it is used |
|---|---|---|
| Written answer forms | 17,723 | Possible answers, solver candidates, and practice puzzles |
| Guess-only forms | 49,623 | Modelled probes that contain a standalone left-side zero |
| All modelled guesses | 67,346 | Candidate questions for the solver |
| Commutative winning groups | 10,757 | Answer forms treated as equivalent under the accepted rule |
A written form is the exact character string. A commutative group can contain more than one accepted arrangement, so 17,723 written forms collapse to 10,757 winning groups when that rule is enabled.
These are NerdleBuddy model counts, not an exhaustive export of the official game. Nerdle says its 17,723 answer calculations were selected from more than 100,000 valid calculations, and its accepted-guess rules can change. See the official Classic Nerdle description.
How Guesses Are Ranked#
For a proposed guess, the solver computes the tile feedback it would receive against every answer still possible. Answers with the same feedback form a bucket. A useful guess produces many small buckets, because almost any reply leaves a short candidate list.
Starter rankings measure full games under NerdleBuddy's generated follow-up strategy. The current published starter artifact stores and sorts the expected number of guesses to four decimal places. That makes the published order reproducible, but equations tied at that precision must be treated as tied rather than assigned a meaningful exact ordinal rank. It also does not prove that no better strategy exists.
The default live solver and player treat a commutative rearrangement as a win. The current difficulty artifact and historical saved decision-tree distribution use a stricter written-form benchmark. They are useful within their labelled scope, but a 2.983 real-rules opener average and a 3.044 strict-tree average are not interchangeable.
The plain-English solver guide explains feedback buckets with examples. The live solver applies the same feedback rules to the rows you enter.
Difficulty Rating#
Difficulty is a percentile rank across all 17,723 modelled written answer forms. Answers are ordered first by the number of guesses used on the saved strict-written-form strategy path, then by the size of their feedback bucket after the strategy's opener. The equation string is a deterministic final tie-break.
The ordered list is split into five equal-sized groups. A 5/5 answer is in the hardest 20% under this benchmark; 1/5 is in the easiest 20%. It is a consistent comparison tool, not a prediction of how any one person will feel about a puzzle or how many guesses the commutative-win player will require.
Practice Puzzle Dates#
NerdleBuddy's archive uses January 1, 2000 as a convenient epoch. Dates walk a seeded permutation of the full answer-form pool, so each form appears once before the next reshuffled loop starts. The same UTC date always resolves to the same practice equation.
This assignment is retroactive. It is a way to organize original practice games, not a record of official Nerdle puzzles. Nerdle itself launched in 2022, and every archive page carries that distinction beside the practice controls.
Classic Daily Answer#
The single Nerdle answer today page is separate from the practice archive. It reads Nerdle's public dated Classic answer page, then verifies the requested UTC date, expected puzzle number, eight-character format, and mathematical equality. Model membership is checked separately; an unsupported equation can still be sourced while its model-derived difficulty and candidate counts are withheld.
The source response is cached for 15 minutes. If the official page is late, unavailable, or changes format, NerdleBuddy shows “being verified.” It never labels yesterday's result as today's and does not decode or publish future schedule entries.
Limitations and Updates#
The equation generator models the published game rules, but Nerdle controls the official game, schedule, and any rule changes. NerdleBuddy's answer pool should not be treated as an official list. Strategy results also depend on the exact pool, feedback implementation, candidate sampling, and saved decision tree.
Dataset files are generated by scripts in the project and checked into the same code base as the site. Editorial update dates change when copy, calculations, or source behavior materially changes—not merely to make a page appear fresh.
Frequently Asked Questions#
Does NerdleBuddy know the official future answer schedule?
The public game client reveals that the schedule is pre-generated, but NerdleBuddy does not use that to publish future answers. The Today page waits for Nerdle's public dated answer page after the puzzle has started.
Is the highest-ranked opener proven globally optimal?
No. It has the best measured full-game result in NerdleBuddy's published ranking under the generated strategy. That is strong evidence for this model, not a proof over every possible strategy.
Are the archive dates official Nerdle history?
No. NerdleBuddy assigns its own generated practice equations to dates retroactively. The archive is a repeatable way to choose a practice puzzle, not a record of official Nerdle games.
Put the method to work
Use your own tile feedback in the solver, or read the strategy explanation.