The case law is linked above, the "Feist Publications, Inc., v. Rural Telephone Service Co" lawsuit that sets some minimum guidelines for what counts as a copyrightable arrangement of facts. And that standard is pretty low, it basically just requires some kind of authorship.
The courts care about amount the method used to create the collection. You are right that if the wordlist had been created by selecting the top ~2000 words from an objective vocab survey or a frequency of use list, then it's unlikely to be eligible.
But the wordle wordlist wasn't created that way, it was "authored". They also filtered out offensive words.
> as it turns out, it was the author's girlfriend who categorized each of the words. Not the author. If there is copyright in the selection (which I doubt), NYT doesn't appear own it.
That's a bold claim. Nobody has seen the wordle sale paperwork, but I'm willing to bet that the lawyers went out of their way to make sure the copyright of the wordlist was assigned to NYT and the creator's partner was fairly compensated.
One of the primary reasons for sale was because the creator didn't want to deal with all the clones, so they would have bought in expensive intellectual property lawyers to make sure the sale was done right.
