This is a Node.js implementation of a Wordle guessing bot. Uses entropy and several heuristics to average 3.514 guesses per answer with 100% accuracy! Tested over 5000 test cases, accuracy is defined as getting the answer within the 6 allowed guesses
This project was built over the course of 4 days so don't expect it to be fast or optimal. It was however an interesting dive into entropy, bitwise optimisation, and implementing the Wordle ruleset!
Key Features:
- Full Wordle ruleset implementation
- Efficient guess culling with heuristics
- Positional letter frequency
- Previous guess overlaps
- Capable of using most* 5 letter word lists (*with ascii a-z/A-Z charset)
- Precomputed feedback matrix with memory mapped reading
- Automated benchmarking capable of thousands of tests
- Feedback matrix consumes significant disk and memory space
- Limited to 5-letter Wordle variants
- Assumes only ASCII characters a-z in wordlist
Interested in the technical details behind this project? Check out how-it-works.md
- Node.js + npm
- Update npm
npm install npm@latest -g
- Run
npm install -g wordulator- Installs base Wordle solutions and guesses
- Precomputes feedback matrix - This step might take a while
- Creates 5000 benchmark test cases from the solution list
- Clone This Repo
git clone https://github.com/ArloFilley/Wordulator
- Navigate to repo folder
cd Wordulator - Create Data Folder
mkdir data
- Install library packages
npm i
- Run
scripts/postinstall.jsnode ./scripts/postinstall.js
- Lanuch the web server and solvers
node . web - Visit localhost:3000 in the browser
 Result.png)
node .
Distributed under the MIT License. See LICENSE.txt for more information.
Arlo Filley — Contact