A quarter of a million real guesses in Borderline paint a clear picture of the world's geography knowledge. 98% of players can find Brazil on a blank globe. 25% can find Timor-Leste. In between sit about a hundred countries that most people half know. You'd recognise the names in a list, but you couldn't point to them. That gap is learnable. Most people just try to close it the wrong way.
The natural way to revise geography is to look at a map. It feels productive and it barely works, because recognising a country on a labelled map is a different skill from recalling it on an empty one. Psychologists have measured this since the 1880s, when Ebbinghaus first mapped how fast passively reviewed information decays. What builds memory that lasts is retrieval: being asked the question and having to produce the answer yourself. In the best known study on this (Roediger and Karpicke, 2006), students who were tested on a text remembered about half as much again a week later as students who spent the same time re-reading it.
Timing is the other lever. Reviewing a fact a minute after learning it does very little. Retrieving the same fact just before you would have forgotten it strengthens it a lot. This is the spacing effect, one of the most replicated results in the field. In practice it means testing yourself soon at first, then after days, then weeks.
Learn mode is a spaced-repetition scheduler underneath a game. Every country you've ever been asked sits on a seven-step ladder, and the rungs are the review gaps:
Miss a country and it returns in 90 seconds. Answer it and the gap grows: 8 minutes, then 45, then a day, three days, a week, three weeks. Miss it later on and it drops two rungs and climbs again. Nail a country first try the very first time it comes up and the game leaves it alone for a day, then quietly checks you weren't fluking.
Two details do most of the work. Everything feeds one memory: countries you miss in the daily or in practice join the same queue, so the drilling aims at your actual weak spots instead of a generic list. And the answer is always a tap on an unlabelled globe, never multiple choice. You practise the exact skill you're trying to build.
Does it work? We can measure it, because every review is a real test. Take the seventeen thousand cases where a player got a country wrong on its first Learn appearance. On the second meeting, the first tap lands inside the right borders 54% of the time. By the fifth, 73%. And those later reviews are no longer cramming: by that point the ladder has spread them days or weeks apart.
Give it a month and the hardest countries on our list stop being hard.
Start with today's five countriesfree, no signup, works offline