This shows an outline of the topics covered in our presentation.

​

Wordle Helpers

Jackson Muller, Dawson Hartman, Andy Zhang, Justin Yu

​

What is Wordle Helpers?

• Wordle is a 5 letter word guessing game

• Get 6 attempts to guess the randomly selected word each day

• Our helpers aim to give the user guess suggestions based on our word processing

• 2 main helpers:

  • Test Your Starter Word

  • Wordle Helper

​

What data are we processing?

Several Lists of 5 letter Words

• NYT Possible Solutions

• NYT Possible Guesses

• Combined List (12,966 words)

​

Frequency Analysis

• Iterates through the whole word list

• Stores frequencies of each letter

• Stores the frequency of each letter in each possible position

• Normalizes the frequency of each letter to compare their relative magnitudes

• Prioritizes guesses that include words of high frequency letters in high frequency positions

​

Cosine Similarity

• compares the length-26 vector of the letter frequency of the full list with a similar length-26 vector for each word

• Essentially ignores magnitude

• Results in a “score” between 0 and 1

• Achieves a result of finding the words that contain the most common letters

• 2 types: positional and nonpositional

  • Positional has added benefit of filtering out uncommon letter positions (plural words)

​

Copula Based Outlier Detection

• Usually meant for identifying the chances of a data point being an outlier or an anomaly of the data set

• In our case the words are the data set and the outliers are the words with less optimal letter frequencies

• We modified it to find the most average word of data set

• Uses marginal distribution functions of the left and right tails and the skew of the data set 

• Finds the best solution by selecting the distribution that minimizes the chances of it being an outlier

​

Greedy Heuristics 

• 3b1b Method

• “Greedy” algorithms take the most optimal choice available at every step

• Goal for Wordle: minimize the resulting possible word lists (plural)

• Done by maximizing the expected value of received information

• Quantified below; more background exists on our website

​

Plans for the Future

• Improve functionality of the website’s built in UI

• Fix/finish our processor in the website to run the code from our analysis

• Identify more edge cases where our algorithms fail/don’t get the word in 6 guesses

• Running simulations to get data on average guesses for each of the algorithms

​

Shortcomings of Wordle Helpers

• All of our algorithms assume Hard Mode which means that you can only guess words that have the green squares

• Guessing a “random” word that uses none of the yellow or green squares to narrow down letter pool isn’t allowed in Hard Mode

• Main fault is when solution word has many rhyming words (i.g. Batch, latch, catch, match)

• Solution is to break Hard Mode which is not something that we want, since the purpose is to solve the word.

​

Demos

1. Testing Your Starter Word

2. Preview of WordleHelpers with full analysis