2.6 KiB
| title | date | tags |
|---|---|---|
| Cheating at Word Games: Part 3 | 2022-01-05T07:50:29-08:00 | [wordle] |
The third in a series on [Wordle]({{< ref "/posts/cheating-at-word-games" >}}).
I created a User Script that tells you how many words you have "narrowed it down to" at each point in the guessing - and, once you've solved the puzzle, tells you what those words were (if there aren't too many to display). I plan to later introduce a feature that tells you the "score" of your guess (calculated as "the inverse of the expected size of the partitions of the answer space", as in my [previous post]({{< ref "/posts/cheating-at-word-games" >}})) and what the optimal guess would have been for even partitioning. Pull requests (and/or comments on my probably-horrendous JavaScript) welcome! In the lovely non-intrusive spirit of the Wordle game itself, the script doesn't "phone home" to measure the number of users - though, if you do end up using it and enjoying it, I'd love for you to let me know!
To install it, you'll need either Greasemonkey (for Firefox) or Tampermonkey (for any other browsers, including Chrome). Install either of these extensions, navigate to the User Script itself, and you should be prompted to install it.
Developing this script really helped me to recognize that my information theoretic viewpoint of "reducing the expected size of partitions" wasn't aligned with how humans solve these puzzles. The key difference is that a human is trying to think of a word, not to select it from an authoritatively-known and enumerable list. For a human there is a difference between "having knowledge in your brain" and "being able to bring it to mind". That is to say - a clue that fixes a letter with a correct evaluation is more useful to a human than one which greatly reduces the partition-space, because that fixed letter will act as a solid prompt that will help the player bring the appropriate word to mind more easily. Maybe I'll develop another scoring strategy that incentivizes correct letters - though it might be subjective as to how heavily they should be weighted.
(Props to @mastergeorge and @NotBrunoAgain for letting me bounce ideas off them during development)