I want to play KEALOA!
That’s not a question. But click here to play anyway.
I found a bug. Now what?
Please send a note to eric@puzzlehead.org and tell me exactly what’s wrong. Tell me which URL, page, tab/subtab, table, column, row, data point – whatever info you can give me to help me narrow it down. I’ll get it fixed ASAP.
What makes a round either True, Near, or Free?
A single round of KEALOA is one of three types based upon commonality of letters and positions among all of the possible answers.
- True: At least one letter in the same position in all words.
- Near: At least one letter in common in all words but not at the same position.
- Free: No letters in common (everything else)
What does Alternation mean?
Alternation measures how often the guess and correct answer changes within a single round. It is defined as the number of actual changes divided by the number of opportunities to change.
In a round of 10 clues, there are 9 opportunities to change.
- If all of the answers were the same, the alternation would be 0/9 or 0%.
- If the answer changed with every clue, the alternation would be 9/9 or 100%.
- If the answer changed 4 times, the alternation would be 4/9 or 44%.
The metric is a variation of the Wald-Wolfowitz runs test.
What does Evenness mean?
Evenness measures how evenly distributed the guesses and correct answers are across the clues.
For example, imagine a round of 10 clues with 2 possible answer words. Here’s how different distributions of those answers across the clues would affect the evenness.
| Count of Word 1 | Count of Word 2 | Evenness |
| 10 | 0 | 0% (completely uneven) |
| 9 | 1 | 47% |
| 8 | 2 | 72% |
| 5 | 5 | 100% (completely even) |
The metric is a variation of the Pielou Evenness Index, which directly normalizes the Shannon Diversity Index, which measures species evenness.
The results look a bit weird (you’d think 9:1 would be closer to 0% than 47%), but the Pielou approach yields a better measure of evenness for this purpose than the alternatives. It also works for rounds that have more than two possible answers.
Where did you get the data?
I listened to every single round of KEALOA, transcribing and verifying each one manually, assisted by the defunct website XWordInfo. (I tried using AI transcription, but the results were … unimpressive.)
I’ve also listened to every episode of Fill Me In, so this makes me a KEALOA doppel compleatist.
Can I get the data?
Sure. Just use the REST-API provided by this site. (If you don’t know what that means, look it up.)
What do you mean that this site was “vibed by Claude”?
The part of this site that manages the KEALOA data is a WordPress plugin that I created using the Claude Code AI agent. I didn’t write a single line of code … I literally just told Claude in plain English sentences what I wanted, and he made it. It was, as they say in my profession, “a highly iterative process” … I’d tell Claude to make something, he’d make it, I’d deploy it, I’d test it, while testing I’d think of something new, and lather-rinse-repeat.
It is genuinely freaky to see just how good these coding bots have become. If I had tried to code it by hand, it would have taken me months, and it would have been a mess … assuming I even finished it at all.
Can I get the KEALOA plugin code?
Sure – it’s at github.com/epeterso2/kealoa-reference. It’s covered by the Creative Commons BY-NC-SA 4.0 license, so give me credit, don’t charge for it, and share your mods with other people.