This is a market-making round of the kind Jane Street runs at the superday: you quote a two-sided market — a bid and an ask — on an uncertain quantity, and the counterparty trades against your quotes. A common version uses the minimum of several dice with different face counts. After each trade a die is revealed and you re-quote, so the round tests whether your fair value actually updates with new information, not just whether you can compute one expectation at the start.
This free game reproduces that structure. Three dice (D24, D64, D128) are rolled hidden and you make a market on the min. Five rounds escalate the way interviewers do: the D128 and D64 faces trickle in before each re-quote, the counterparty switches from random fills to picking off only your mispriced side, the maximum spread gets capped, and the final rounds add a sealed-bid auction to see the hidden D24 early. Every quote is scored against the exact conditional fair value, so you can see where your numbers drifted.
A market-making round, typically at the superday for trading roles. The interviewer asks you to quote a bid and an ask on an uncertain quantity (often dice-based), trades against your quotes, reveals information, and expects you to re-quote. You are judged on fair-value accuracy, spread discipline, and how you react to being traded against.
Condition on what is revealed. With nothing revealed, compute E[min] over all three distributions. Once the D64 and D128 faces are known with smaller value m, fair value is E[min(D24, m)] = (m(m+1)/2 + (24−m)·m) / 24 for m < 24 — only the unrevealed D24 still carries uncertainty.
A defensible fair value, a spread that matches your actual uncertainty, fast updates when information arrives, and awareness of adverse selection — if the counterparty keeps hitting only one side of your market, your quote is probably off. Recovering calmly after a losing trade counts more than any single fill.