Jeremy B. Merrill

Purple really is the hardest color in New York Times's Connections game.

Across the 400 games that I analyzed to see how similar each group's words are, yellow is easiest on average, then green is a little harder, then blue, then purple is hardest.

You can read more about my analysis of Connections here: https://jeremybmerrill.com/blog/2024/08/this-algorithm-solves-nyt-connections.html

Is purple the hardest category for you in Connections? Did you even realize that the colors had a meaning? (I didn't until I did this analysis.)

How well can an algorithm solve the NYT Connections game?

to be hidden

Aug 29, 2024 at 1:16amWeb
Show threadJ. Nathan Matias 🦣Aug 29, 2024
@jeremybmerrill your last footnote <3 sigh
Show threadJeremy B. MerrillAug 29, 2024
@natematias this is not the most impactful or important consequence of that, but it's just a bummer the trivial, fun knowledge we'll never get to know
Show threadMarcos HuertaAug 29, 2024

@jeremybmerrill Did you think about trying other embeddings besides Glove? Like sentence transformers? Maybe even Llama or other locally runable LLMs? I wouldn’t trust the chatbot answer but the embeddings might be helpful for computing similarity.

I have been totally nerdsniped. I am going to go down my own rabbit hole on this momentarily.

Show threadragrumAug 29, 2024
@jeremybmerrill that's cool that the stats perfectly back up the trickiness scale that the setter aims for
Show threadJeremy B. MerrillAug 29, 2024
@ragrum oh interesting, I hadn't seen that! Thank you