Forget chasing “vibes” with AUDIO or CRANE. A new study from Binghamton University and MIT researchers reveals the best Wordle strategy isn’t guessing the most probable word — it’s choosing guesses that eliminate the most uncertainty. Using Shannon entropy, the math of information theory, their solver wins 99% of the time and solves most puzzles in 3–4 guesses.
The Core Insight: Information Beats Intuition
Led by Assistant Professor Congyu “Peter” Wu at Binghamton’s Watson College, the team applied Shannon entropy to measure how much each guess reduces uncertainty. “A guess doesn’t have to be the most likely answer; it simply has to be informative,” said doctoral student Donald Stephens. The goal shifts from “being right” to “maximizing expected reduction in uncertainty”.
In practice, that means opening with words like LARES, RALES, TARES, SOARE, SLANE, or SALET — not because they’re common answers, but because they reveal the most about the hidden word. The strategy can look random because it prioritizes data gain over directly solving.
How It Performs: 99%+ Win Rate, 3.42 Average Guesses
Across simulations of all 2,315 Wordle solution words, the entropy-based method solved 99.9% of puzzles within six tries. MIT’s Exact Dynamic Programming model, published in Operations Research, found SALET solves the game in 3.421 guesses on average — never needing the sixth guess. It wins in two guesses 4% of the time and three guesses 57% of the time.
For comparison, letter-frequency tactics and “best first word” lists like AUDIO average 3.51+ guesses and succeed ∼90% of the time.
Top Starting Words From Research
- MIT Sloan/Ops Research: SALET — obsolete form of “salett,” an engraved helmet. Reveals 1.683 colored tiles on average vs 1.320 for AUDIO.
- Binghamton/Information Theory: LARES, RALES, TARES, SOARE, REAIS top the list.
- Economic Times simulations: ADEPT, CLAMP, PLAID give 98.79% success if playing optimally.
- Hard Mode guarantee: Northeastern researchers found SCAMP or SCOWL as first guess guarantees a win in ≤5 moves.
Two-word combos matter too. SLANE + an optimal second word yields 10.04 bits of information, cutting 2,309 possibilities to just 2.19 words on average.
How to Use It: Run a Script, Enter Feedback
To play like the researchers, you’d enter your guess results into a separate program after each turn. It reads the gray/yellow/green feedback and recommends the next word that provides the most useful information. Coders have posted open-source solvers on GitHub, though many warn it “takes the fun away”.
Why It Works: Wordle Is a Feedback System
Every guess shrinks the search space. Green = right letter, right spot. Yellow = right letter, wrong spot. Gray = not in word. The entropy approach picks words that split the remaining word list as evenly as possible, guaranteeing the fastest path to the answer — even if the guess itself isn’t a possible solution.
