Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/fkodom/wordle

Fastest Wordle solver in the West.
https://github.com/fkodom/wordle

game python wordle wordle-assistant wordle-helper wordle-solver

Last synced: 7 days ago
JSON representation

Fastest Wordle solver in the West.

Host: GitHub
URL: https://github.com/fkodom/wordle
Owner: fkodom
License: mit
Created: 2022-01-30T14:11:50.000Z (almost 3 years ago)
Default Branch: main
Last Pushed: 2022-03-06T15:14:08.000Z (almost 3 years ago)
Last Synced: 2024-10-28T06:30:02.935Z (about 2 months ago)
Topics: game, python, wordle, wordle-assistant, wordle-helper, wordle-solver
Language: Python
Homepage:
Size: 1.52 MB
Stars: 2
Watchers: 1
Forks: 0
Open Issues: 0
Metadata Files:
- Readme: README.md
- License: LICENSE

Awesome Lists containing this project

README

# wordle

Fastest [Wordle](https://www.powerlanguage.co.uk/wordle/) solver in the West.

* 99.83% success rate
* 3.597 average turns to win
* <10 ms per guess (single-threaded)
* **Measured against the official Wordle word list!** (Always check your sources. 🙃)

Use the `--mode win-percentage` flag to optimize for success rate (see below). In that case, you can't lose if your first guess is RALPH. ¯\\_(ツ)_/¯

![quick draw](https://media.giphy.com/media/aqDXCH2M1ycEw/giphy.gif)

## Install

Install directly from this repository:
```bash
pip install git+https://github.com/fkodom/wordle.git
```

## Solve

Launch the assistive solver:
```bash
solve-wordle
```

Or optimize for win percentage:
```bash
solve-wordle --mode win-percentage
```

## Play

Visit the **[Public Web App](https://share.streamlit.io/fkodom/wordle/main/app.py)**, or play a command line game:
```bash
play-wordle
```

## Benchmarks

Full details in [benchmarks.jsonl](data/benchmarks.jsonl).

* Accuracy (99.83%) and turns-to-win (3.597) are averages over recommended starting words, using default settings
* RALPH has the highest win percentage (99.95%)
* SLATE has the fewest turns to win (3.539)
* BLAST is a good balanced choice (99.91%, 3.608)

**NOTE:** RALPH has a 100% win percentage if you use the `--mode win-percentage` flag.

## How It Works

Exactly solving for word probabilities requires an exhaustive search through all possible word combinations. (There are way too many to be fast or practical.) Instead, we approximate them using a cheaper method.

### Maximum Split

Each guess should eliminate as many possible words from the word bank as possible. The "maximum split" method does just that:
* iterate through each possible `(guess, target)` pair
* measure the number of remaining possible words for each pair
* recommend guesses that give the smallest average numbers of remaining words.

### Word Probability

A few definitions:

> $\bg_white p_i(c)$ --> probability of character $c$ at position $i$ in the word
>
> $\bg_white p(c)$ --> probability of character $c$ at any position
>
> $\bg_white p(w)$ --> probability of a word ($c_1$, $c_2$, $c_3$, $c_4$, $c_5$)
>
> $\bg_white n_c(w)$ --> counts of character $c$ in word $w$
>
> $\bg_white n_c$ --> counts of character $c$ in the remaining word bank

Very roughly speaking, the word probability scales with each of $p_i(c)$ and $p(c)$:

$\bg_white p(w) \sim \prod_i p_i(c_i) \, p(c_i)$

Then, $\bg_white p_i(c)$ and $\bg_white p(c)$ can be approximated from the remaining possible words, just by counting the frequencies of different letters. Then, to avoid over-counting from words with repeated letters, we divide by the number of identical permutations.

$\bg_white p(c) \sim \frac{n_{c}}{n_{all} \cdot n_c(w)!}$

### Hybrid (Default)

The "maximum splits" method is more accurate, but it's slow when the number of remaining words is large. So as a hybrid method:
* If >128 possible words remain, use **word probability**
* Otherwise, use **maximum split**

The `--mode win-percentage` flag uses exhaustive search once the number of remaining words drops below 16. It's somewhat of a hack, but you win a slightly higher percentage of games. Unless RALPH is your starting word, it's probably too small of a difference to notice.