There's a math-optimal Wordle strategy. The best starting word. The elimination logic. Here's everything from information theory applied to a five-letter word game.
I've played Wordle every single day since early 2022. That's over 1,400 consecutive puzzles. My average solve is 3.2 guesses, and I've never lost a streak longer than 200 days. I'm not saying this to brag — I'm saying it because Wordle isn't about vocabulary. It's about information theory, and once you understand the math, the game becomes a very different experience.
Most people treat Wordle as a word game. It's not. It's a search algorithm with a human interface. Every guess you make should maximize the information you extract from the result. Every green, yellow, and gray tile is a constraint that shrinks the remaining solution space. The question isn't "what word can I think of?" — it's "what guess eliminates the most possibilities?"
Let me break down everything I've learned from over four years of daily play, a lot of reading about information theory, and way too many spreadsheets.
Before we get into strategy, it's worth understanding why this specific game became a cultural phenomenon. There have been word-guessing games for decades. Mastermind (the board game) uses the exact same colored-peg feedback mechanic with colors instead of letters. Jotto, a pen-and-paper game from the 1950s, is literally Wordle without the grid.
So why did Wordle explode in late 2021?
Three reasons:
One puzzle per day. This is the single most important design decision. By limiting everyone to one puzzle, the game created artificial scarcity. You couldn't binge it. You couldn't get bored of it. You solved today's puzzle and then you waited. That constraint turned a forgettable word game into a daily ritual.
Shared experience. Because everyone solves the same word on the same day, Wordle became social. You could text your friend "I got today's in 2" without spoiling anything. The colored square grid (more on that later) became a universal language. Suddenly, a single-player word game had the dynamics of a multiplayer event.
No app, no account, no monetization. Josh Wardle built it for his partner. It was a website. No login, no ads, no push notifications begging you to come back. In a world of predatory mobile game design, Wordle felt refreshingly honest. (The New York Times acquisition later changed some of this, but the core simplicity remains.)
The combination created something rare: a game that non-gamers played. My mother-in-law plays Wordle. My dentist plays Wordle. People who haven't touched a video game since Tetris play Wordle every morning with their coffee. That's extraordinary reach for a game with no marketing budget.
Here's where it gets interesting. Wordle has a fixed answer list of roughly 2,300 words (the original list — the NYT has modified it slightly over time). Your job is to identify one word from that list in six guesses or fewer.
Each guess gives you feedback: green (right letter, right position), yellow (right letter, wrong position), or gray (letter not in the word). This feedback is information — it tells you something about the answer that you didn't know before.
Information theory, developed by Claude Shannon in 1948, gives us a way to quantify this. The key concept is entropy, measured in bits. If you have 2,300 possible words, the entropy is about 11.2 bits (log₂ of 2,300). To identify the answer, you need to extract roughly 11.2 bits of information from your guesses.
Each guess can theoretically provide up to about 5-6 bits of information if it produces a pattern that roughly halves the remaining possibilities. In practice, some guesses are more informative than others.
The expected information of a guess is calculated by looking at every possible answer, determining what feedback pattern you'd get, grouping the remaining possibilities by pattern, and computing the entropy reduction. A guess that produces many distinct feedback patterns (spreading the remaining words into many small groups) is more informative than one that produces few patterns (leaving you with a few large groups).
This is why "PIZZA" is a terrible first guess even though it's a real word. The double Z and the relatively rare letters mean most answers will produce the same gray-gray-gray-gray pattern, telling you almost nothing. Meanwhile, a word like "SALET" produces dozens of distinct patterns, each one dramatically shrinking your search space.
Everyone has opinions about the best starting word. "ADIEU because it has four vowels." "AUDIO for the same reason." "CRANE because the NYT bot recommended it."
Opinions are fine. Data is better.
When you compute the expected information for every valid five-letter word against the full answer list, here are the top performers:
Tier 1 — Maximum Information (4.8+ bits):
Tier 2 — Excellent (4.7+ bits):
Notice a pattern? The best starters are heavy on R, S, T, A, E, L — the most frequent letters in five-letter English words. They avoid uncommon letters (J, Q, X, Z) and they spread their letters across different positions where those letters are most likely to appear.
SALET (a type of medieval helmet) consistently tops computational analyses. It's not a word most people know, and that's fine — you're not trying to guess the answer on move one. You're trying to learn as much as possible.
If using an obscure word bothers you, SLATE, CRATE, or CRANE are all within 0.1 bits of optimal. The difference is negligible in practice. Pick one that feels natural to you and use it consistently. Consistency matters more than optimality because it lets you build pattern recognition around a familiar starting position.
My personal starter is SLATE. It's a common word, it's easy to remember, and it gives me S, L, A, T, E — five of the six most common letters in the answer list.
Here's something most strategy guides get wrong: they obsess over the starting word and then say "now just think of words." The second guess is where the real skill lives.
After your first guess, you have feedback. Maybe SLATE gave you a green A in position 3 and a yellow T. Now you know the answer has A in position 3, T somewhere (not position 4), and doesn't have S, L, or E.
The naive approach: think of a word with A in position 3 and T somewhere else. MATCH? BATCH? WATCH?
The optimal approach: think about what guess would give you the most new information. You already know about S, L, A, T, E. Your second guess should test different letters while respecting the constraints you've learned.
This is why many strong Wordle players use a two-word opening sequence — a fixed first and second word that together cover 10 unique letters. Popular pairs:
After two well-chosen guesses covering 10 letters, you've tested almost half the alphabet. You typically have enough constraints to narrow the answer to a handful of words — sometimes just one.
The tradeoff: using a fixed second word ignores the information from your first guess. If SLATE gives you three greens, you should absolutely NOT play CRONY — you should guess the answer. Fixed pairs work best when your first guess gives you minimal information (all grays or one yellow).
My rule of thumb: if the first guess gives me 0-1 greens and 0-1 yellows, I play my fixed second word. If it gives me 2+ greens or 2+ yellows, I switch to targeted guessing.
Wordle's Hard Mode requires you to use all confirmed information in every subsequent guess. If you got a green S in position 1, every future guess must start with S. If you got a yellow R, every future guess must contain R.
Normal Mode advantages:
Hard Mode advantages:
The dirty secret is that Hard Mode barely affects the average solve count for skilled players. The mathematical difference is roughly 0.1-0.2 additional guesses on average. Where Hard Mode really hurts is on specific tricky words — words like WATCH/MATCH/BATCH/CATCH/HATCH/LATCH/PATCH where six words share four letters and you can only test one per guess in Hard Mode.
In Normal Mode, you could guess something like "CHAMP" to test C, H, M, and P simultaneously, even though CHAMP can't be the answer. In Hard Mode, you're stuck guessing _ATCH words one by one and potentially running out of guesses.
My recommendation: play Normal Mode if you care about your stats. Play Hard Mode if you want the extra challenge and don't mind the occasional loss on a trap word.
Not all letters are created equal in Wordle. Here's the frequency of each letter across the ~2,300-word answer list:
Top Tier (appearing in 30%+ of answers): E (46%), A (39%), R (37%), O (29%), T (29%), I (28%), S (28%), L (27%)
Mid Tier (15-27%): N (24%), U (20%), C (17%), Y (17%), H (16%), D (16%), P (15%)
Low Tier (under 15%): G (12%), M (12%), B (11%), K (8%), F (8%), W (8%), V (5%), X (2%), Z (2%), J (1%), Q (1%)
This is why your first guess should target the top tier. Testing E, A, R, S, T in your first word means you're checking letters that collectively appear in the vast majority of answers. Testing J, Q, X, Z would be absurd — those four letters combined appear in barely 6% of answers.
But frequency alone isn't enough. You also need to think about...
The letter E appears in 46% of Wordle answers, but it's not evenly distributed across positions:
E overwhelmingly favors positions 4 and 5. So a word with E in position 5 is more informative than one with E in position 1, because it's testing E where E is most likely to actually be.
Some other notable position preferences:
S loves position 1 (25% of all S appearances) — words like STARE, STONE, SHALE.
Y almost always appears in position 5 (82% of Y appearances) — WORDY, HAPPY, GRAVY. If you get a yellow Y, immediately move it to position 5.
A is most common in position 2 (36%) and position 3 (27%) — TAKEN, MAPLE, CHARM.
R is fairly evenly distributed but slightly prefers position 3 and position 4.
T strongly favors position 1 (30%) — TIGER, TOAST, TRAIN. Getting a yellow T should make you consider moving it to position 1.
This positional awareness is what separates good Wordle players from great ones. When you get a yellow letter, don't just include it in your next guess — put it where it statistically wants to be.
English words aren't random. They follow patterns, and recognizing these patterns lets you generate candidate words much faster. Here are the most important ones for Wordle:
The -IGHT family: EIGHT, FIGHT, LIGHT, MIGHT, NIGHT, RIGHT, SIGHT, TIGHT, WIGHT. If you confirm _IGHT early, you're in trouble in Hard Mode — that's nine possibilities with only one distinguishing letter.
The -OUND family: BOUND, FOUND, HOUND, MOUND, POUND, ROUND, SOUND, WOUND. Eight words, same deal.
The -ATCH family: BATCH, CATCH, HATCH, LATCH, MATCH, PATCH, WATCH. Seven words.
The -ASTE family: BASTE, HASTE, PASTE, TASTE, WASTE. Five words.
Double letters: This is the single biggest trap in Wordle. About 25% of answers contain a repeated letter. Words like SPEED (double E), LLAMA — wait, that's not five letters. Words like CREEP, FLOOD, SPELL, STIFF, DWELL. Many players instinctively avoid double letters, which means they'll never guess these words until they've eliminated all single-letter options.
My advice: don't avoid double letters, but don't prioritize them either. If your constraints point toward a double-letter word, go for it. Common doubles to watch for: EE (CREEP, STEEL), LL (DWELL, SKILL), SS (GROSS, PRESS), OO (FLOOD, BROOD), TT (LATTE, KITTY).
Consonant clusters: Some consonant combinations are extremely common in five-letter words:
Recognizing these clusters helps you construct candidate words faster. If you know the word has R in position 2 and a vowel in position 3, the answer probably starts with a consonant cluster: CR_, DR_, FR_, GR_, PR_, TR_.
Here's my actual solving process, step by step:
Step 1: Play your starter. I play SLATE. Every day, same word, no exceptions.
Step 2: Count your constraints. After SLATE, I catalog: which letters are confirmed (green), which letters are present but misplaced (yellow), and which letters are eliminated (gray). I also note which positions are determined.
Step 3: Estimate the remaining possibilities. This is where experience helps. If I have 2 greens and a yellow, I know I'm down to maybe 5-15 words. If I have all grays, I might still be looking at 100+ possibilities.
Step 4: Decide — eliminate or guess?
Step 5: After guess 2, get aggressive. By guess 3, you should have enough information to start guessing the answer directly. If you're still playing elimination words on guess 4, something went wrong earlier.
Step 6: On guess 5 or 6, always guess the answer. Never waste a late guess on elimination. If you have three possibilities left and two guesses, you need to guess answers — you can't afford to eliminate.
Let me walk through a concrete example:
Guess 1: SLATE → All gray. No S, L, A, T, E in the answer.
This is actually great information. We've eliminated five of the most common letters. The answer uses uncommon letters. I immediately think of letters NOT yet tested from the mid-tier: N, U, C, Y, H, D, P.
Guess 2: CORNY → C gray, O yellow, R green (position 3), N yellow, Y gray.
Now I know: has O (not position 2), has R (position 3), has N (not position 4), no C, no Y. Combined with guess 1: no S, L, A, T, E, C, Y.
The word has ?R? with O and N somewhere, using only B, D, F, G, H, I, J, K, M, O, P, Q, U, V, W, X, Z for the remaining letters.
Words fitting: DRINK? No, that would need... wait. _IR_N? GIRON? No. What about PRION? P-R-I-O-N. R in position 3? No, that's position 2.
Let me reconsider. Position 3 is R. So R. Need O and N, not in positions 2 and 4.
UNWORN? Too many letters. FURROW? Six letters.
IMBRO... no. THROW? No, T is eliminated.
UNBORN? Six letters. ADORN? A is eliminated. ACORN? A is eliminated.
GROIN? No, that's G-R-O-I-N. R is position 2, not 3.
FUROR? _U-R-O-R. That has R in position 3 and 5, O in position 4... but N isn't included.
DIRON? Not a word. BIPOD? Missing O and N placement.
MINOR. M-I-N-O-R. N in position 3? No, that's 3=N, but we need 3=R.
VIGOR? No N. RIGOR? No N, and R in position 1 and 5.
WORLD? No, L eliminated.
WHORL? L eliminated.
GOURD? G-O-U-R-D. R in position 4, not 3.
MORON? M-O-R-O-N. R in position 3! O in positions 2 and 4. N in position 5. But O can't be in position 2 (yellow in position 2 from guess 2). Wait — actually the yellow means O IS in the word but NOT in position 2. So O in position 4 is fine. But does MORON have double O? M-O-R-O-N. Yes. O in position 2 is still there though, and we said O isn't in position 2.
Hmm, MORON has O in position 2 AND position 4. Since we know O is NOT in position 2, MORON doesn't work.
What about NIRON? Not a word. MYRRH? No O or N.
IRONY? I-R-O-N-Y. Y is eliminated.
NURON? Not a word. GIRON? Not a standard Wordle answer.
DURNO? Not a word. BURN? Four letters.
PRONG? No, that's P-R-O-N-G. R in position 2.
WRONG. W-R-O-N-G. R in position 2, not 3. Nope.
HERON? No, E is eliminated.
APRON? A is eliminated.
DROWN? D-R-O-W-N. R in position 2, not 3.
DIRON, GIRON... what about ENDOW? No, we need R in position 3.
Let me think differently. R where all of S, L, A, T, E, C, Y are excluded.
UNRIG? Not likely a Wordle word. FROWN? F-R-O-W-N. R in position 2.
I'll be honest — this is genuinely hard. Which is the whole point. Sometimes the constraints lead you to a corner where the answer is an uncommon word, and that's where your vocabulary gets tested.
Guess 3: ROBIN → R gray (wait, R was green before)...
Actually, let me abandon this example before it gets more confusing. The point is: the process works even when the specific answer is tricky. Systematic elimination will always outperform random guessing.
This is the single most important decision you make in every game, and it comes down to a simple expected-value calculation.
Guess (go for the answer) when:
Eliminate (test new letters) when:
The -ATCH trap is the classic example of when elimination beats guessing. Suppose after two guesses you know the answer is _ATCH. In Hard Mode, you're stuck guessing BATCH, CATCH, HATCH, etc. — each guess only eliminates one word. You might need 5-6 guesses.
In Normal Mode, you can play a word like CLUMP to test C (CATCH), L (LATCH), M (MATCH), and P (PATCH) simultaneously. One elimination guess cuts your remaining possibilities from seven to two or three.
This is the mathematically correct play, even though it "wastes" a guess on a word that can't possibly be the answer. You're trading one guaranteed non-answer for dramatically more information.
Let's talk about the sharing format for a moment, because it's genuinely fascinating from a design perspective.
AkousaWordle 1,276 3/6
⬛⬛🟨⬛⬛
🟩⬛🟩🟨⬛
🟩🟩🟩🟩🟩
This grid is pure genius. It communicates your performance (3 guesses), your journey (how the greens and yellows accumulated), and absolutely nothing about the actual answer. You can share it the moment you finish without spoiling anyone.
The format spread across Twitter, WhatsApp groups, Slack channels, and family group chats. People who never posted about games were posting colored squares. It created a daily micro-competition without the toxicity of a leaderboard. You weren't competing against anyone — you were sharing an experience.
I've seen marriages improved by Wordle sharing (morning ritual together), offices bonded by it (Slack channel dedicated to daily results), and genuine friendships formed over it (strangers comparing strategies). For a game with no multiplayer, no chat, and no social features, that's remarkable.
The success of Wordle spawned an ecosystem of variants that take the core mechanic in different directions:
Dordle — Solve two Wordles simultaneously. Your guesses apply to both boards. This is a meaningful step up in difficulty because your guesses need to be efficient across two different words. You get 7 guesses instead of 6.
Quordle — Four Wordles at once. Nine guesses. This is where strategy really starts to diverge from standard Wordle. You need to balance information gathering across four boards, and a guess that's great for one board might be useless for another.
Octordle — Eight Wordles. Thirteen guesses. At this point, your first 2-3 guesses should be purely about letter coverage, not about any specific board. The two-word opening strategy becomes essentially mandatory.
Sedecordle — Sixteen Wordles. Twenty-one guesses. This is the endurance test. You'll spend more time reading the boards than thinking about guesses.
Nerdle — Wordle but with math equations instead of words. You guess an eight-character equation (like 9*8-7=65) and get the same green/yellow/gray feedback. It tests a completely different set of skills — number sense and equation construction rather than vocabulary.
Worldle — Guess the country from its silhouette. After each guess, you're told how far away (in kilometers) and in what direction the answer is. Geography nerds love this one.
Heardle — Guess the song from increasingly long audio clips. (RIP — Spotify shut this one down, but clones exist.)
Each variant emphasizes different aspects of the core Wordle skill set. Dordle and Quordle reward efficient letter coverage. Nerdle rewards mathematical thinking. Worldle is purely knowledge-based. If you've plateaued at standard Wordle, variants offer fresh challenge without learning a new game.
There's something psychologically powerful about a daily puzzle. It's a small, completable task with clear success criteria. In a world where most of our work is ambiguous — "is this code good enough?" "did that meeting accomplish anything?" — Wordle gives you an unambiguous win every morning.
The streak counter amplifies this. Once you have a 50-day streak, losing feels genuinely painful. At 200 days, it's unthinkable. The streak becomes a source of quiet pride, a tiny monument to consistency.
This isn't unique to Wordle — Duolingo, meditation apps, and habit trackers all exploit the same psychology. But Wordle does it more honestly. There's no premium tier, no streak freeze you can buy, no notification spam. You either played today or you didn't.
My advice for building a Wordle habit:
The original Wordle's one-per-day limit is beautiful for habit building but frustrating when you want to practice strategy. Fortunately, there are ways to play unlimited rounds.
Several sites offer Wordle clones with random words — you can play as many rounds as you want. Some even let you configure the word length (4-letter, 6-letter, 7-letter variants) or the word list (common words only vs. all valid words).
This unlimited practice is genuinely useful for improving. It lets you:
I'd recommend doing 10-20 unlimited rounds per week if you're serious about improving. Focus on consistency — track your average guess count over 50+ rounds to see if a strategy change actually helps.
Wordle popularized the "daily word puzzle" format, but it's just one species in a rich ecosystem of word games. Here's how the major types compare:
Wordle-style (deduction): You have hidden information and use logical elimination to find it. Skills: pattern matching, letter frequency intuition, constraint satisfaction. Appeal: satisfying "aha" moments, shareable results.
Crosswords (knowledge + wordplay): Fill a grid using intersecting clues. Skills: vocabulary, general knowledge, understanding of crossword conventions (abbreviations, themes). Appeal: deep, meditative solving experience. Much longer sessions.
Word Search (visual scanning): Find hidden words in a letter grid. Skills: visual pattern recognition, spatial awareness. Appeal: relaxing, low-pressure, good for winding down.
Anagram games (recombination): Rearrange letters to form words. Skills: mental flexibility, vocabulary breadth. Appeal: fast-paced, competitive potential.
Spelling Bee / word construction: Make words from a limited set of letters. Skills: vocabulary depth, systematic exploration. Appeal: hunt for rare words, satisfaction of finding the pangram.
Hangman (classic guessing): Guess letters to reveal a hidden word. Skills: letter frequency knowledge, vocabulary. Appeal: simple, social, works for all ages.
Each type exercises slightly different cognitive skills. Wordle is primarily about logical deduction. Crosswords test knowledge breadth. Word searches are about perception. If you enjoy Wordle, you'll probably also enjoy crosswords and spelling bees — they share the "constrained search" mechanic that makes deduction games satisfying.
Here's a technique I use that I haven't seen written up elsewhere. After each guess, I mentally construct a probability heatmap of which letters are likely in which positions.
For example, suppose after guess 1, I know:
I weight each possibility by letter frequency in that position. This gives me a fuzzy picture of the answer's shape, and I look for words that match the highest-probability cells.
This is essentially what computational solvers do, but approximated by human intuition. You won't be as precise as an algorithm, but you'll be much faster and more systematic than someone guessing randomly.
After four years, these are the rules I swear by:
There's something I find deeply satisfying about Wordle that goes beyond the game itself. It's a daily reminder that complex problems can be solved systematically. That information, carefully gathered and logically applied, can cut through uncertainty. That patience and consistency beat cleverness and luck.
In a world that feels increasingly chaotic, spending three minutes each morning methodically reducing 2,300 possibilities to one feels like a small act of sanity.
If you're looking to expand beyond your daily Wordle fix — whether that's variants like Quordle and Nerdle, classic crosswords, competitive word searches, or entirely different word games — there are plenty of word games out there that scratch the same itch in different ways. The important thing is to keep that daily puzzle habit alive. Your brain will thank you.