I spoiled Wordle with this one weird trick

I wouldn’t describe myself as competitive, exactly. It’s just that whenever there’s an opportunity to do so, I absolutely must prove that I’m more clever than anybody else. So when I saw people on twitter claiming to have an infallible new strategy for solving the daily Wordle, I had to prove them wrong. Impossible! I thought. If there truly was such a strategy, I’d have thought of it myself.

Here’s the gist, if you haven’t been on the internet recently: Wordle asks you to guess a five-letter word within six attempts, each attempt giving you clues about the letters you got right and wrong. “Green” means you got the letter exactly right, “Yellow” means the letter is in the word in a different position, and black means the letter does not appear in the word. Basic strategy is to guess a word at random, look at the clues, and then guess subsequent words which match the clues you’ve been given. If the clues tell you that the first letter is “b” and there’s a “d” in there somewhere, you might guess “bread”, “bodes” or “bride”, for example. Repeat until the options narrow down to one. So far, so straightforward. That strategy has stood me in pretty good stead, usually getting the answer within 4 or 5 guesses, with the occasional 3-guess triumph.

But here’s where the ridiculous strategy comes in. Instead of guessing a word that matches the clues you’ve been given, you guess a COMPLETELY DIFFERENT word. A word that uses all-new letters that you haven’t used before. The idea is that this strategy will give you more information more quickly than repeatedly guessing similar words. Even with almost every letter correct, you can easily spend three guesses on “fires”, “tires” and “mires” before finding out that the correct answer was “wires”.

But this is nonsense, surely? You have six guesses, and a good score is only three or four guesses. Why would you waste a precious guess on a word that has no chance of being correct? Those coveted two or three-guess scores are only possible with a bit of luck. Why would you make a guess that makes it impossible to get a lucky win?

I had to prove these fools wrong, so I turned to code to bash together a quick solver. This would give me incontrovertible, repeatable proof of the effectiveness of my preferred strategy. I started with a simple approach which I called the “Random Solver”. The random solver guesses a word at random, looks at the clues for that word, and then guesses a random word that matches the clues its been given. It repeats that until it guesses the word right or it runs out of guesses.

The random solver is not very good at solving Wordles. Only about 65% of puzzles are solved within 4 guesses, and only around 95% are solved at all. You can most likely do as well or better yourself.

To do better, we need our solver to make better guesses. The first way to achieve this is to stop making guesses at random, but rather to guess words that have more common letters in them, and to avoid guessing words with double letters in them if possible. These two improvements mean that the solver will eliminate more possibilities more quickly. It’s extremely effective. What I ended up calling the “Smart Solver” can solve 76% of puzzles within 4 guesses, and can solve 98% before running out of guesses. That’s impressive. I’d wager that most people would struggle to match those numbers. It’s not looking promising for the ridiculous strategy.

So how did the “Strategy Solver” do? I coded it up. I set it so that if there were more than half a dozen possible words it could guess, it would instead try to eliminate letters by guessing a word with a set of common letters that it hadn’t used before. Then I ran it against the same 500 Wordles I’d used to test my previous solvers. And then I got very, very annoyed. It won! Despite basically never getting the right word on the second guess (as the other solvers did in 2 or 3% of cases), it can solve 78% of puzzles within 4 or fewer guesses, and solves 99% before running out of guesses. It’s a small margin, but it’s consistently better than the approach I expected to easily win. What’s worse is that when I looked at what choices it made, it turned out that it was guessing the same two words pretty much every time. Not only was this strategy better than the approach I had been using, it was also substantially easier to use.

To avoid spoiling Wordle for other people, I won’t tell you what those two words are, but let me assure you that I have not solved a Wordle in fewer than four guesses since I discovered them, and the whole game has become substantially less enjoyable for me. I guess the real lesson for me here is that sometimes it is possible to be too clever for your own good.