A deep analysis of Wordle by Xan Gregg
Created:
Xan (@XanGregg) has been playing Wordle since the start of this year. You can see that his stats, at right, are good. Most times he has played, he has needed four guesses. And he has never failed to guess the word. (Alas, I can't say that about myself! I've missed it once.)
In January, Xan wrote a post called "Deep Wordle" on his personal blog, Raw Data Studies, about his analysis of Wordle. He described his post as "a deep traversal of the forking paths of guesses that make up the full decision tree." The bottom line: His Wordle analysis found that most any start word is good. He did find a few start words that might land you in trouble.
Xan and his Wordle analysis were featured this week in an article by journalist Mona Chalabi for The Guardian titled "What the numbers tell us about the best – and worst – Wordle start words." I asked Xan to share his Wordle strategy and findings from his analysis.
What's your approach to playing Wordle?
Xan: Since my analysis found that any start word is good, I use a different start word each day when I play. I start with the previous day's solution when I can remember it. It's a bit of a handicap when I have to start with a less informative word like "knoll," but it's worked out OK so far.
What’s the gist of your deep Wordle analysis that you want Wordle fans to know about?
Xan: When Wordle became popular, many best word analyses sprang up that I would characterize as "shallow" in that they would only go one or two levels deep into the tree of possible guesses. With a "deep" analysis, I hoped to determine the maximum of total guesses needed with optimal play. It turns out that for the 3,600 start words I analyzed, all but three need a maximum of five guesses to reach any solution. Those three are "mummy," "shush" and "yummy."
What do you mean by 'optimal play'?
Xan: For my analysis, it means that each guess after the first one is chosen to maximize the chance of solving in a total of five guesses. I have an example at the end of my "Deep Wordle" post; it gets pretty deep in the details.
How did Mona Chalabi find you, and what did she ask you?
Xan: She contacted me about getting some data from my analysis for an idea she was working on. Not sure how she found my analysis. Ultimately, the data I had wasn't great for her idea since the optimization process tends to result in short games without many interesting patterns.
Do you play any of the other Wordle spinoffs like Worldle?
Xan: I like to play Quordle. The idea is to solve four Wordles simultaneously using the same guesses. I start with the same three words each day and then try to solve each in turn for a total of seven guesses. Not sure if it's always possible -- that may be my next analysis topic!
Anything else you want to add?
Xan: "Jazzy" is one of the worst start words for Wordle that I investigated. But with optimal play, it will still lead to any solution in five total guesses.
Comments