Old Mission Capital Interview Guide

How the Old Mission Capital interview process works

For trading applicants, the reported route is application → timed assessment → recruiter or trader interview → deeper technical or desk discussion. For the 2026 campus Software Engineering track, the first test may be in person instead. The work reflects Old Mission’s trading business: its teams price and provide liquidity in listed securities, with researchers developing models, engineers building trading systems and traders managing risk. The firm grew from pricing illiquid ETFs before expanding across exchange-traded asset classes.

1. Apply to the appropriate role. Quant Trader, SWE and Quant Research each need a different technical preparation plan. Old Mission recruits both graduates and experienced candidates.
2. Complete the first assessment. Trading reports describe timed maths plus Python; 2026 campus SWE reports describe a proctored fundamentals test on paper.
3. Explain your decisions live. Trading rounds include probability and interactive quotes; engineering rounds involve programming language behaviour, algorithms and systems; research discussions involve models and evidence.
4. Meet the team or desk. Reported later interviews go deeper into trade-offs, your judgement and how you respond when the problem changes.

The key distinction is the actual task. A junior trader can pass through a mathematical speed test and then struggle with managing an inventory in a trading game. A software applicant can be challenged on memory layout or Linux without seeing a standard LeetCode puzzle. The sections below prepare you for both.

Old Mission Capital online assessment: questions, timing and format

The most recent reported trading OA combines probability or sequence questions with a coding task. A June 2026 junior trader account describes 25 quantitative problems in 35 minutes, then one coding problem in 25 minutes. The maths was manageable but speed mattered; the programming task required hash maps and careful string formatting. Another September 2026 Chicago junior trader account names HackerRank, probability, sequences and a final Python question.

Older accounts differ: a 2023 new-grad report describes 35 maths questions in 25 minutes with a coding task, while a 2024 trader report recalls about 35 maths questions and a running-median task without specifying the time limit. These are separate recruiting-cycle observations, not additional sections of the June 2026 test.

What are the trading maths questions actually like?

Expect short problems where a correct setup matters more than advanced theory: count possible outcomes, update a probability after new information, calculate an expected payoff or recognise the next term in a sequence. The coding portion deserves separate attention. The June 2026 report specifically mentions dictionary logic and output-format mistakes; a 2024 trader report describes maintaining a running median. Practise both simple stateful Python tasks and careful edge-case testing rather than spending all your time on calculus.

Technical interviews and final rounds, by role

The biggest differences appear after the first test. Choose your track below for the question families, reported interview tasks and the most useful preparation.

Old Mission-style practice questions, with worked solutions

These are original EverythingQuant practice exercises designed around the skills above. They are not described as questions taken from Old Mission's actual interviews.

1. Trading probability: two independent signals

An ETF's fair value has a 40% chance of being above its quoted price. Two independent analysts each report HIGH with 75% probability when it is above and 25% when it is not. The reports are independent conditional on the true state. If both report HIGH, what is the probability that fair value is above the quote?

2. Market making: a fill changes your position

An ETF is initially worth 100 to you based on its constituent basket. You quote 98 bid / 102 ask. A counterparty buys one share from you at 102. New information lifts your fair value estimate to 106. What is your inventory and marked P&L?

3. SWE fundamentals: the Python aliasing trap

What does this function return for [10, 20, 30], and how would you fix it if you wanted three separate buckets?

def distribute(values):
    buckets = [[]] * 3
    for i, value in enumerate(values):
        buckets[i % 3].append(value)
    return buckets

4. Coding stretch: lower running median

After each incoming price, return the lower median of all prices so far (for an even number of observations, take the smaller of the two middle values). For [5, 1, 4, 2], the answer is [5, 1, 4, 2]. A pair of heaps handles a stream without sorting every prefix.

import heapq

def lower_medians(values):
    lower, upper, result = [], [], []
    for value in values:
        heapq.heappush(lower, -value)
        heapq.heappush(upper, -heapq.heappop(lower))
        if len(upper) > len(lower):
            heapq.heappush(lower, -heapq.heappop(upper))
        result.append(-lower[0])
    return result

Here lower is a max-heap represented by negated values and upper is a min-heap. They stay ordered and the lower heap holds the extra item when the count is odd. Each insertion costs O(log n); processing n values costs O(n log n) time and O(n) space. This is a harder extension of the running-median problem family described by a 2024 trading candidate.

5. Research judgement: is the backtest persuasive?

A daily trading signal was selected after testing many candidates. On its best 100 trading days, its mean return is 4 basis points, with a standard deviation of 20 basis points. Assuming independent daily returns for a first calculation, what is its approximate t-statistic, and what would you check before trusting it?

How difficult are Old Mission Capital interviews?

The biggest hurdle for junior traders is speed, then trading judgement. Both recent quant OA accounts call the core mathematical questions approachable but time-sensitive. Later market-making interviews are tougher because you must update estimates while keeping a correct position and P&L. The 2025 floor-trader report specifically identifies market making as the stage that decided the candidate's outcome.

For SWE, the difficulty is depth rather than just LeetCode rank. In 2026, candidates describe questions about memory and C++ behaviour that go below the usual university coding exercises. Python-focused roles still test actual Python semantics, Linux and implementation. Research applicants should expect to justify how their models work under imperfect market data rather than merely recite formulae.

How to prepare: a focused one-week plan

Spend the week on the exact task your role is likely to present. The schedule below is an EverythingQuant preparation plan, not an employer-specified syllabus.

For the trader track, start with EverythingQuant's probability questions and Mental Maths and Sequences, then practise updating live quotes in the market games. If conditional probability and expectation are weak, work through Probability for Quants. For SWE, add the DSA course and programming problems.

Old Mission Capital interview FAQs