blossom solver

Blossom Solver

A solver for the Blossom word game. The solver is for non-commercial education purposes only. All trandemarks belong to Merriam-Webster.

Try it

https://kensnyder.github.io/blossom-solver/

Motivation

I love playing Blossom but always wonder how my score stacks up against the best possible score. This tool attempts to find a high score for a given puzzle. You can also use it in "Hints mode" before you complete the puzzle. It will give you hints including the number of possible pangrams, number of possible words, high score, and first letters of the top words.

Dictionaries

Overview

Merriam-Webster does not specify what dictionary is used for the game. We only have some guidelines:

1. No hyphenated words
2. No swear words
3. No proper nouns
4. No medical words

Additionally, due to the nature of the game, we have the following constraints:

1. Words must have 7 or fewer distinct letters
2. Words must have 4 or more letters

Process

Many dictionaries are available online. The Scrabble dictionary is perhaps the best guess at the contents of the Blossom dictionary, but it does not contain words with more than 15 letters because those are impossible to play in Scrabble.

Levels

This Blossom solver includes 3 different dictionaries for 3 different skill levels.

Level 3 (Expert)

I started with the largest dictionary I could find, from this GitHub repository: https://github.com/dwyl/english-words which has about 370k words.

Note that it and all other files mentioned below are in the data directory of this repository.

1. That dictionary, saved as data/full-dictionary.txt, contains about 370k words.
2. I removed all words with 4 or fewer letters and 7 or fewer distinct letters.
3. I then broke that file into smaller files by number of letters, which are found in data/*-letters.txt
4. In chunks, I checked the HTML for each dictionary entry from Merriam-Webster's website.
5. That allowed me to identify medical words, proper nouns, and words not in Merriam-Webster's dictionary. Those are saved in data/medical-words.txt, data/proper-nouns.txt, and data/not-in-dictionary.txt.
6. I then downloaded a list of swear words from this GitHub Repository. and saved it rot13-encoded as data/swear-words.rot13.txt
7. I then used scripts/compile-words.mjs to remove all those invalid words from data/full-dictionary.txt, leaving 118k words saved to data/uncompiled-level3.txt.

Level 2 (Intermediate)

1. I started with the "Wiktionary 100k", a 2005 compilation from Wiktionary.org. That file is saved as data/wiktionary-100k.txt.
2. I removed all words with 4 or fewer letters and 7 or fewer distinct letters.
3. I removed all words not present in the level 3 dictionary
4. I saved the result to data/uncompiled-level2.txt. That file contains 26k words.

Level 1 (Beginner)

Coming soon. I will likely use the Level 2 dictionary and remove words that I've personally never heard of.

But wait!

After you complete a game, Blossom will show you all the words and pangrams that were possible. Looking at words on the list, I discovered some words that in the Blossom game that are not in data/full-dictionary.txt. I manually added those words to data/discovered-words.txt. Those are included in the final uncompiled and compiled dictionaries. Similarly, words I found in our dictionary but not in Blossom are saved to data/discovered-non-words.txt.

Beyond the dictionary - which words go with which bonus petal letter?

I used letter frequencies in each dictionary to order the alphabet from least common letter to most common letter. The bot uses score then letter frequency to choose its final plays. It is more naive than permutation or some mathematical algorithm, but usually gets the maximum score or very close to it.

Precompilation

I've chosen to precompile the dictionaries to allow the solver to provide solutions in a matter of milliseconds. The downside is that your browser has to download these larger files upon page load.

By example, here is the scheme for the compilation format:

mustard24+
muster11
mustered27mustrd32e+

1. Words start at a line break and continue until encountering a number.
2. A plus is added to the end of the line if it is a pangram.
3. If the word has multiple sections of letters/numbers, the sections indicate how many points each bonus petal letter would produce.

Example breakdown:

# mustard is a pangram worth 24 points regardless of which letter is the bonus petal
mustard24+

# muster is not a pangram and worth 11 points regardless of which letter is the bonus petal.
muster11

# mustered is a pangram
# If the bonus petal is m, u, s, t, r or d, it is worth 27 points
# If the bonus petal is e, it is worth 32 points
mustered27mustrd32e+

License

ISC License (ISC)