What it does: sets how many places from the current scope are in the rotation; everything else sits on the bench.
Why it's useful: a small active set (around 10) repeats each place often enough to stick. Master those, then grow the set or let the rules below rotate in fresh material. Shrinking benches your most-answered places first (all-time); growing brings back your highest-weight benched places, so the set stays on what you know least. You can also bench or activate individual places with the switches in the table below.

What it does: once you answer a place correctly this many times in a row in the current session, it drops out of the active set. A wrong answer or a peek resets its streak to zero.
Why it's useful: it clears away the places you've mastered so every remaining question lands where it counts. The active set shrinks toward zero as you work through it, a natural way to "finish" a scope in one sitting.

What it does: when a place's weight decays below the cutoff (you've answered it right enough times to count as mastered), it's swapped out for the highest-weight benched place still at or above the cutoff. The active set keeps its size and cycles in fresh material, shrinking only when no benched place qualifies.
Why it's useful: it's the "endless drill" counterpart to auto-retire: work through a whole continent a handful of places at a time without touching the controls. Lower cutoffs demand more proof of mastery before a swap; higher cutoffs rotate faster.

What it does: marks an answer right when it is spelled the way the name sounds, or with letters doubled or single where the real name has it the other way round: Tailand, Kirgizstan, Misisipi, Filipines, Wagadugu, Arkansaw. It never blurs two different places together: anything that could just as easily be somewhere else (Deli is both Delhi and Dili) still comes back as a spelling check, and typing one country's name when another was asked is wrong at any setting.
Why it's useful: it keeps a geography quiz about geography rather than orthography, which is the point when you are learning names you have only ever heard. Quizzes taken with it on rank on the Leaderboards and keep a spelling score: the share of your correct answers that needed no forgiving. Boards break ties by score, then spelling, then time, so leaning on the forgiveness costs placement against a run that spelled everything, never the score itself.

Hold the microphone to talk swaps the hot microphone (open for the whole question, hands free) for push-to-talk: it opens only while you hold the microphone button down and closes the moment you let go, which is the one to pick in a room with other people talking. Letting go keeps what you said rather than discarding it, so with Send it automatically on as well, releasing the button is what sends the answer.
What it does: the answer box listens instead of waiting to be typed. What you say appears as you say it, tinted while the microphone is still deciding and settling at each pause, and several goes build one answer up (Saint Kitts … and Nevis). Read it, then send it: swipe the box right, press Submit, or press Enter. Wrong word? Swipe left, press the ✕ or press Esc, and it clears and starts listening again straight away, even if the microphone hasn't finished thinking. Typing anything hands the box back to you until the next question. Switching it on the second way, the microphone inside the box, works on every tab.
Send it automatically is off by default because reading the answer back before it counts is most of the point; with it on, a settled answer sends itself after the wait you set here (up to three seconds, in quarter-second steps, and the amber bar under the box draws exactly that long), and anything that clears the box calls it off: saying another word, the ✕, a left swipe, or typing. Set the wait to 0s and there is no window at all: the answer goes the instant the microphone settles on it, which is the fastest way through a run and also means a name you say in two breaths (Saint Kitts … and Nevis) will be sent after the first one.
Worth knowing: a microphone gives back dictionary spellings of what it thought it heard, and guesses phonetically at names it has never met, so this pairs well with Forgiving spelling above. Nothing about a spoken answer is scored differently from a typed one. Where the browser can recognize speech on-device it is asked to, and the audio never leaves the machine; otherwise the browser itself sends it away to be transcribed, which is the same thing that happens when you dictate anywhere else in it.

This browser can't recognize speech, so voice answers are unavailable here. Chrome, Edge and Safari can; Firefox can't yet, and on iPhone or iPad it works in Safari but not in a copy installed to the Home Screen.

What it does: a soft mark in the ear when an answer is graded, in Practice and the Quiz: two short notes stepping up for right, one low note bending down for wrong. Deliberately quiet and a fraction of a second each, so a long session never grates, and synthesized on the spot (this site ships no sound files). Turn it off for a silent site; Versus has its own sounds and its own mute.