What Jane Street and Citadel test when finance knowledge is optional

Quantitative recruiting has accumulated a mythology of impossible brainteasers, instant mental arithmetic, and secret market knowledge. The official guidance from Jane Street and Citadel is more specific and, in some ways, more demanding. Both firms want to observe how a candidate works through uncertainty with another person.

Jane Street is a trading firm. Citadel is an investment firm, while Citadel Securities is a market maker. Their roles and interview processes are not interchangeable. Still, their public materials expose a shared selection signal: strong quantitative foundations matter, but so do questions, assumptions, explanation, and the ability to change course.

Jane Street removes finance knowledge as the hiding place

Jane Street's official interview page says quantitative trading candidates are not tested on finance or economics knowledge. It calls finance background optional and describes the interview as collaborative problem solving. For strategy and product roles, it similarly emphasizes how quickly candidates learn and approach unfamiliar problems rather than prior frameworks.

Removing specialized market knowledge does not make the interview easy. It makes prior exposure less useful as camouflage. The interviewer can watch whether you identify missing information, represent uncertainty, notice an incorrect assumption, and build on a hint.

Jane Street also publishes a Probability and Markets Guide. That is an important distinction: finance trivia may be optional, but quantitative fluency is not. Basic probability, expected value, conditional reasoning, and comfort with games of chance create the language in which the collaborative problem can happen.

Citadel expects foundations before finance

Citadel's quantitative research materials say market expertise can be learned on the job, but robust statistical intuition should already exist. The firm highlights mathematics, statistics, coding, working with data, and applied research. Its interview guide advises candidates to assess the problem before coding, ask clarifying questions, share their reasoning, discuss tradeoffs, and acknowledge what they do not know.

That combination matters. A candidate can be technically fast and still be a weak researcher if they rush past the problem definition. Another can be mathematically sophisticated but difficult to collaborate with if every assumption remains private. Quantitative work turns a messy world into a model, and every model leaves something out. The interview needs to reveal whether the candidate understands what was excluded.

Citadel also describes research as a loop that connects ideas, tests, code, deployment, and market feedback. For students, the relevant lesson is not to trade. It is to practice making a prediction, specifying why it could be wrong, and updating after evidence arrives.

Five abilities hidden inside a quant problem

An original practice problem

Consider this exercise, created for practice rather than taken from any firm's interview bank:

A good response begins by restating the reinforcement rule and separating the two possible first draws. If the first token is blue, the second-draw bag has four blue and two gold tokens. If the first is gold, it has three blue and three gold. The final probability combines those branches using the probability of each first draw.

The numerical answer matters, but the visible structure matters more. Can the student explain why the second draw is conditional on the first? Can they check the result with a short simulation? Can they predict how the answer changes if five tokens are drawn under the same reinforcement rule?

Build a quantitative reasoning notebook

The most useful practice record is not a page of correct answers. It is a sequence of forecasts, models, checks, and updates. Give each entry the same structure:

1. Question: State the decision or quantity before opening the dataset.
2. Initial estimate: Record a prediction and a plausible range. Do not edit it after seeing the result.
3. Assumptions: List the facts being treated as stable and the conditions that would break the model.
4. Method: Explain the probability model, simulation, regression, or other analysis in plain language.
5. Result: Show the calculation or code, then translate the output into a sentence.
6. Update: Compare the result with the prediction and identify the assumption that deserves the most attention next.

Four drills that build the right reflexes

• Probability games: Use cards, dice, and tokens to create small games. Solve the expected value, then simulate thousands of plays and explain any gap.
• Estimation with ranges: Estimate a quantity such as daily meals served in a school district. State low, central, and high cases instead of hiding uncertainty behind one number.
• Prediction calibration: Make weekly yes-or-no predictions with confidence levels, then review whether events labeled 70 percent actually occur about seven times in ten over a sufficiently large set.
• Small data investigations: Ask one question of a public or synthetic dataset, reserve part of the data for checking, and explain why a pattern might fail outside the sample.

Common preparation traps

• Memorizing leaked questions. Recognition disappears as soon as one condition changes, and it prevents genuine collaboration.
• Equating speed with strength. Fast arithmetic is useful, but a quick answer to the wrong model is still wrong.
• Hiding uncertainty. Pretending to know is more damaging than naming the gap and proposing how to resolve it.
• Overfitting a backtest. Trying enough variations will eventually produce an attractive historical chart. Without a held-out test and a clear prior hypothesis, that chart may be evidence of search rather than insight.
• Practicing only alone. The official materials emphasize interaction. At least half of practice should include explaining, receiving a hint, and revising in real time.

The deeper selection signal is calibration

Probability problems are not only math tests. They expose whether a person can hold a belief with the right amount of confidence. Applied research adds another layer: can that belief survive data, criticism, and changing conditions?

Students can build that ability without a Bloomberg terminal or a brokerage account. Ask precise questions. Predict before measuring. Explain the model. Invite someone to challenge it. Update without embarrassment. Those habits are useful in quant interviews because they are useful in quantitative work. They also connect directly to the broader framework in the Selection Signals synthesis.