A round of about five minutes on the one piece of arithmetic almost everyone under-reads: what an amount turns into when each year's interest joins the balance. Nine cards arrive one at a time — a one-off amount left alone, a payment made every year, a start put off by a decade, a fee taken annually, three paying-in plans racing to the same finish line. You plant a flag before the snowball rolls, and then it rolls. Within 8% is a hit, within 20% is half credit, and a miss costs one of three chances. Every rate on every card is hypothetical and stated as such, and you never type a figure of your own.
The goal
Nine cards in three stages of three, getting harder as they go. Five kinds appear: the finish value of a one-off amount; the total after paying in at the start of every year; what share is lost by starting some years late; what share a yearly fee takes over decades; and a three-lane race in which you pick the plan that finishes furthest. Every card's three slots name the condition that card turns on — the amount, the yearly payment, the years started late or the yearly fee — followed by the hypothetical annual rate and the horizon, and every card compounds once a year — which is printed above the track, because with the frequency unstated a card would have several correct answers.
Controls and play
Estimate cards take a number: tap the on-screen keypad or type, then press Enter or the plant-flag button. There is no dragging while you answer — the only slider in the game is the sandbox on the result page, and it has minus and plus buttons and left and right arrow keys that do exactly the same thing. Race cards take one of three lanes: press 1, 2 or 3, or Tab to a lane and press Enter. Text that cannot be read as a number, and a number outside the allowed range, are not answers: the button stays off and the clock keeps running. Only the clock running out marks a card wrong on your behalf.
Points and streaks
A hit — within 8% of the real finish, or within 1.5 percentage points on a share card, whichever is wider — scores full credit. Close, within 20% or 4 points, scores half and keeps your streak alive. A miss scores nothing, resets the streak and costs one of three chances. Cards are weighted 1 at the first tier, 1.3 at the second and 1.6 at the third, and the multiplier rises half a step every three non-misses up to three times. There is no speed bonus at all: paying for speed is paying for guessing. The grade runs on a weighted quality ratio out of all nine cards, so stopping early never improves it.
Finishing and results
The round ends on the ninth card or when the third chance goes. The result page puts every flag you planted against the real finish on one log scale, so you can see at a glance whether you landed behind the curve every time — the headline counts only the estimate cards you actually answered, and says separately how many you never reached. Below it: a card-by-card review, two knowledge cards that are the same in both languages, and a sandbox where an amount, a hypothetical rate and a horizon can be moved between preset steps to watch the curve respond. The sandbox scores nothing and takes no figure of yours.
Replay and progress
Every round is dealt from a seed and the share text carries it: opening that link deals the same nine cards, in the same order, with the same answers — and in either language, because the deal is language-neutral and only the sentences differ. "Play again" draws a new seed and drops the old link from the address bar. Leaving mid-round keeps the card you were on, the score, the streak, the chances and the seconds that were left, counted against that card's own limit — thirty seconds on a race card, not forty. Both clocks stop while the page is hidden.
Language and context
Both languages play the identical round: the same nine cards, the same parameters and the same answers, because nothing in the arithmetic depends on where you live. Only the unit differs — thousands of dollars in English, 万元 in Chinese — and that is a label on the input row, not a conversion. The result page's two knowledge cards are the same in both languages for the same reason. Where local rules genuinely differ, such as how often a bank applies interest, the game says that frequency is an assumption rather than teaching one market's practice as the rule.
