Using data from a quarter-million tweets to turn Wordle into a numbers game, and what it teaches us about the truth.
I would imagine that very few people tweet results when they get stumped.
Something I'd love to test about "hard mode" is how many legitimate English words remain after certain green letters. It'd depend on how deeply the game pool dips into the obscure--I'd hope not very far, since that'd be a frustrating experience. I agree that fishing for more individual letters is less useful than reducing the pool of possible words down to a handful. And when you focus on words that might fit, the heuristic-based pattern-matching language parts of your brain engage, which are perhaps better suited to a game like this than quantitative, analytical brain regions. Now I go consider how to get as many letters of RSTLNE in a five-letter word as possible...
I played 299 Wordle puzzles before Devang was paid to shut down, and had a winning percentage over 98%. I played "hard mode" about 50% of the time, and found it much easier to win, although more likely to require 4-5 tries. I will miss Wordle from my life, but my average time per game was well below 10 minutes. One game per day isn't worth worrying about. I won't pay the NYT for the privilege. And FWIW, it was the numerous times when I had four greens after two guesses that made me the craziest, because there were often more possibilities than opportunities. I had to try to outguess the potential cultural biases of the algorithm. It gave me more sympathy for those who complained about the SAT's cultural bias, although Wordle is simply a hunt for a real word without context.
I play in mixed-modes I guess. Mostly pinning, but strategically when there is little choice.
My proudest moment:
Cornered! Is it hatch, watch, latch, batch, or patch??? At this point I could get it in 1 guess or 5 guesses… Let’s test them all in one guess, and guarantee a win in 2!
whelp 🟩🟨⬜️⬜️⬜️ Bingo!
Not to bring Bayes into this... but I think the "hard mode" approach forces people to think with a bayesian approach, which helps them in the long run. Once they get green, they update their priors and can reevaluate the probability of each letter appearing in each remaining square. This isn't to say you can't do this in strategic mode, but I think it's a harder, and people are probably doing anxiously doing this before meetings, or ya know, at midnight...