March 17, 2026
Maps, math, and mild chaos
I Simulated 38,612 Countryle Games to Find the Best Strategy
He simulated 38,612 geography games — now everyone’s fighting over the “best” first guess
TLDR: A tinkerer simulated 38,612 rounds of the Countryle geography game and built a solver that averages 2.85 guesses. The crowd split: fans cheered the over-the-top analysis, while skeptics nitpicked the “best” claim and asked for visualizations—turning a daily puzzle into a playful showdown over what “optimal” really means.
The internet is loving and roasting a mega-nerdy deep dive into Countryle, a daily geography puzzle where you guess a country and get hints like north/south, same continent, and whether it’s hotter/colder or more/less populated. One commenter, st0ffregen, claims they built a bot that solves it in 2.85 guesses on average, using “entropy” (simple version: your next guess tries to split the remaining options as evenly as possible) — and people are buzzing.
Cue the drama: Is it really the “best” strategy? User Tepix calls the headline out as wrong, saying there’s still room to improve. The pedants vs. the cheerleaders are in full swing: some want mathematical proof of “optimal,” others are like, relax, the bot is fast and that’s cool. Meanwhile, bubblerme celebrates the glorious overkill, comparing it to those viral Wordle solvers. The vibe? Peak nerd joy.
There’s also a delightful side quest: totetsu wonders what the game’s “state space” would look like if graphed like a sliding-puzzle and drops a link to a Klotski visualization. Commenters toss around puns, “entropy” jokes, and “I brought a PhD to a casual game” memes. Bottom line: a smart, fair (no peeking!) solver ignites the age-old internet debate — what counts as “best,” and does it even matter when you’re just trying to beat today’s map?
Key Points
- •The author simulated 38,612 Countryle games to study optimal strategies using only in-game feedback.
- •Countryle provides five feedback channels: direction, continent, hemisphere, population, and average temperature, with population/temperature bucketed as wrong/close/correct.
- •A modular solver was built with per-signal modules that first filter inconsistent countries and then score the next guess.
- •Shannon entropy is used to select guesses that most evenly split remaining outcomes, maximizing expected information gain.
- •Not all feedback is equally informative; hemisphere is often one-time useful, while direction, temperature, and population typically provide more granular guidance.