Robot that can solve the Daily Wordle Automatically!
As I'm sure many of you know, wordle is a game where you try to guess a random 5-letter word with up to six guesses. It is actually very simple to cheat and find the word right away if you do a little digging around in inspect element, but that's no fun. Instead, this robot can serve as a benchmark to see how good your word deducing skills are. There are two versions of this robot. The "smart" version has a word-bank of words that can only be used as the answer key within wordle. This word-bank is stored as "supposedWords.csv". The "dumb" version uses a much larger word-bank (stored in "validWords.csv") which contains all valid entries you can input into wordle. The dumb version has about 6x more words than the smart version and every word within the smart version exists in the dumb version. I got these lists by looking at the network tab of inspect element on chrome and downloading the only js file that was sent from the server. A copy of this file is saved as "wordle.js" and you can scroll through it to find the lists for yourself if you'd like.
Must have selenium installed version == 4.1.0 chromium version == 97.0.4692
Simply run the main.py file to start the robot. Comment out which word-bank you would like to use.
Selenium starts up and then inputs the initial starting word into the wordle. From there, we do some simple html parsing to see what letters are present, absent, or correct. If any letter is absent, we remove all words with that letter from our dictionary. If a letter is present that implies that it is in the word we are looking for but not at that position. In this case we remove all words that don't contain this letter and all words that contain that letter in that position. If a letter is correct we can remove all words that don't contain that letter in that position. This drastically reduces the word-bank even after 2 attempts. To decide which word to use next, each word has a value associated with it that is derived from the frequency of the letter within a dictionary. This means words with more e's and a's are likely to be guessed next instead of words with j's and w's. Once the answer is guessed, the json is updated with the result and the wordle display is printed in the terminal.
Currently using the tutorial from this video to set up my AWS Lambda environment. https://github.com/soumilshah1995/Selenium-on-AWS-Lambda-Python3.7/blob/main/test.py
- Tweet out daily wordle summary for each bot
- ^Add this to Lambda Function
- Update word frequency calculation to reduce the value of words that have multiple of the same letter
Surprisingly, all of the things mentioned in the Future Expansion section actually got completed. Recently the NYT did change the HTML behind wordle and it took me quite a few days to notice. The most recent code works with the updated HTML and the twitter bot is up and running again! If you ever want to check it out on twitter you can find it @wordlebotsolver.