In a group mock trading game, a contract settles at a company's September scooter sales in thousands. Your opening estimate is 80, range 70 to 90. Clues arrive: registrations running 8% above last year, then a three-day factory shutdown costing about 9,000 units. Set the market, update it after each clue, and manage a 50-lot limit.

Open at 76 bid, 84 offered; move to 79 at 85 after the registrations clue, fair about 82; then to 71 at 75 after the shutdown, fair about 73. Move the mid by what each clue is worth, narrow the width only as uncertainty falls, and quote sizes that leave room under 50 lots for the next surprise. The fills here take you short 10, long 10 and back to flat, and settlement at 73 makes 10 points times lots.

Centre on your estimate and set the width from your uncertainty. A shopkeeper who has never sold a product quotes a wider gap between buying and selling price than one who sells it daily. With a range of 70 to 90, a market of 76 at 84 is centred on 80 and wide enough that one bad fill does not hurt much, while still being tight enough that others will trade with you. Size the quote at 10 lots, not 20: two fills on the same side would still leave you at 20 lots, well inside the 50 limit. The first fill comes at once: someone buys 10 at 84, and you are short 10.

Translate each clue into units before touching the quote. Registrations 8% above last September's 76 thousand point to about 82.1 thousand, so the mid moves up about 2. Registrations are a reasonable proxy for sales, but not identical, so you narrow only a little: 79 at 85, width 6. The shutdown is a direct subtraction: about 9 thousand units off 82.1 leaves about 73.1, so the mid drops nine points at once and, with two pieces of hard information in, the width can fall to 4: 71 at 75. Moving the mid by the value of the clue, and the width by how much it cut uncertainty, is the rule the interviewers are listening for.

After the first clue someone sells you 20 at 79, two points under your new fair value. A fill that arrives just after you move is a signal: the seller may have heard about the shutdown, and you should shade your bid and keep size small until you learn more. You are now long 10 at an average that looks poor once the shutdown clue lands. After it you quote 71 at 75, a buyer takes 10 at 75, and you are flat. Settlement at 73 scores the three trades: short 10 from 84 makes +110, long 20 from 79 loses 120, short 10 from 75 makes +20, a total of +10 points times lots.

Interviewers score the process more than the P&L. They listen for a stated fair value and width before every quote, a sentence on what each clue is worth in units, and a reaction to fills that treats them as information. The limit is a tool, not a ceiling to hit: quoting sizes that keep two adverse fills inside it is what lets you keep making markets when the next clue arrives. The weak game is the opposite: a fixed width all the way through, a mid that jumps on gut feel, and a position that hits 50 lots just before the clue that moves against it. The limit of any game is that the clues are clean; a real desk has to judge how much a noisy data point is worth, which is harder.