I like to think of myself as good at Wordle (caring about Wordle is one of my worst personality flaws and I’m embracing it, thank you very much), so I was thrilled to see that the New York Times has a new tool called WordleBot that can grade your performance on any past Wordle.
If you’ve played Wordle recently, you can visit the tool in the same browser and it’ll know how you made your guesses—but you can also upload a screenshot of a Wordle you have played in the past. I’m going to introduce you to the bot with some older screenshots, but then we’ll dive into strategy for today’s puzzle, number 293, so either stop reading or go solve it now if you don’t want to see spoilers.
How to use WordleBot
First, solve today’s puzzle, and then visit this link. It will guide you through the process. If you want an analysis of an older puzzle or somebody else’s puzzle, take a screenshot. (After you solve a puzzle, the Wordle stats screen will pop up. Hit the “x” and then screenshot the grid. You can also go back to that page any time after solving, so long as the next puzzle isn’t out yet, and it will remember your guesses.)
Then, swipe through the result screens. The first one gives you an overall score for skill and for luck. The rest step through your guesses, and the bot compares what choices you made to what it would have done.
How good is WordleBot at Wordle?
The bot cheats, by the way. It calculates how well you’re narrowing down the possibilities, but it knows which 2,309 words are in the official solutions list. The English language actually has more like 13,000 five-letter words; many were just excluded to make the game more interesting, like plurals of four-letter words that don’t tend to be on there, for example. Some obscure words were also left out, as well as a bunch of words that I don’t consider very obscure: FUTON will never be a solution, the bot told me, nor will PORKY, although both are allowed as guesses.
On the other hand, at least the bot doesn’t know which puzzles have already been published. When I’m playing, if one of the possibilities is SNOUT but I remember that SNOUT was the solution last week, I can mentally cross it off. The bot has not been given this guidance.
How to learn from WordleBot (spoilers ahead!)
The most exciting thing about WordleBot is not that it will tell you how skillful you are, but that it can give you tips to be better.
Today’s puzzle was on the easier side (I and two of my coworkers all solved it in three guesses.) The bot, smug bastard that it is, says it would have solved it in two. (It starts with CRANE.)
I got an overall skill rating of 93. It liked my opener (STONE), and said my second guess of CHAMP cut the possibilities down to just two. My third guess, then, was lucky.
I ran staff writer Sarah Showfety’s solution through the bot next. It gave her an overall skill score of just 85, mainly because her second guess of SNARE only brought the solution pool down to three possibilities instead of two. Deputy editor Joel Cunningham got a skill score of 84, and in this case he seems to have been dinged for being too lucky. His opener brought the possible solutions to just 9, and after CLAMP he was down to just one. (We all beat the typical Wordle player, the bot says: The average skill score is 79.)
Just for fun, I tried today’s puzzle again in another browser and tried my best to play badly. (As we learned from my post about Antiwordle earlier this week, I may not actually be great at Wordle, but I am definitely bad at being bad at it.) It liked my starter, BOGUS, and my follow-up, PARSE. But after that, the little bot was just shaking its head at my bad guesses. I actually had it narrowed down to just four possibilities at that point. My next two, RAISE and SPARE, it said were “wasted” guesses that didn’t add to what we already know. But that’s not actually true: Both gave clues about the position of the letters that were previously yellow, and both included gray letters that I could eliminate as possibilities. I ended up with a skill score of just 22 though, solving the puzzle in five guesses, which fits.
So what did we learn? Among other things, we got our starters graded. CRANE is the bot’s pick; it liked Joel’s STERN and Sarah’s ROUTE. All of these will on average narrow down the pool to somewhere between 90 to 100 possibilities. (That number can be higher or lower, though, depending on what the day’s actual solution is. Sometimes you’re just unlucky.)
So go ahead and ask the bot how it thinks you did. You might be mad about the answers, but you’ll probably learn something.