Hedge Funds interview preparation

Questions

100

Traced to a firm

39

Firms

16

Updated

September 2026

Asked at

All firmsMan Group10Balyasny Asset Management7Bridgewater Associates3DED.E. Shaw3Apollo Global Management2KKR2Oaktree Capital Management2Point722SCSquarepoint Capital2ACAQR Capital Management1BGBaupost Group1Coatue Management1HPS Investment Partners1Northern Trust1Viking Global Investors1Wolverine Trading1

Topic

All topicsStrategy taxonomy8Stock pitch10Short selling6Portfolio construction8Risk and drawdown8Performance and alpha7Event-driven and merger arb8Distressed and credit5Fund structure and economics7Financing, NAV and operations6Compliance and research process5Quant and systematic6India and Category III AIFs5Career and fit11

Level

AnyCoreIntermediateHard

Type

AnyTechnicalMarket viewBrainteaserCaseFit

1. 031How do you understand portfolio risk?Portfolio constructionIntermediatetechnicalMan GroupInvestment Management · Boston · 2022

   Say this

   As three separate questions, not one number. How much do I expect to lose in a normal month, what happens in a bad one, and what am I unknowingly concentrated in? Volatility answers the first, stress tests answer the second, and factor decomposition answers the third.

   Then walk it

   1. Layer one, the normal case: volatility, VaR and contribution to risk per position. Useful for sizing and for spotting that one position is carrying a third of the risk.
   2. Layer two, the bad case: stress tests and scenarios. Rerun the book through 2008, March 2020, the 2021 momentum unwind, a 100 basis point rate shock. This is where you learn the hedges stop working.
   3. Layer three, the hidden case: factor and thematic decomposition. What is the book's net exposure to growth, to momentum, to oil, to the dollar, to one supply chain? Most surprises are a concentration nobody had named.
   4. Then liquidity risk, which sits underneath all of it. Days to exit at 20 percent of volume, and what the book looks like if you have to raise 20 percent of cash in a week. Illiquidity converts a paper loss into a realised one.
   5. And correlation instability, which is the one that actually hurts. Correlations rise in stress, so a diversified book is less diversified precisely when it matters. I would assume correlations go to one in the tail rather than trusting the historical matrix.
   6. The limitation to volunteer: every number here is backward looking and conditional on a covariance matrix estimated from a period that may not resemble the next one. That is why hard limits and drawdown stops exist alongside the models, rather than instead of them.

   Where candidates lose it

   Answering only with VaR or only with volatility. A single risk number is the wrong shape of answer to this question. Name the three layers, then add liquidity and correlation instability, and say explicitly that the models are backward looking. That last admission is what a risk-focused interviewer is listening for.

   Expect next

   • What does VaR miss?
   • How would you stress test a long-short equity book?
   • How do you think about transaction cost?

   Reported by candidates at Man Group (Investment Management, Boston, 2022). Source: Wall Street Oasis.

2. 032What is capacity, and how do you know a strategy is crowded?Portfolio constructionIntermediatesuperdayQuantitative hedge fundsMulti-manager platforms

   Say this

   Capacity is the amount of capital a strategy can run before its own trading destroys the edge. Crowding is the same problem caused by other people: too many funds holding the same positions, so the exit is narrow. Both show up as rising cost and correlated drawdowns rather than as a signal that stops predicting.

   Then walk it

   1. Capacity is a cost problem first. As size grows, each rebalance moves the price more, so realised return falls even though the signal is unchanged. The practical test is to plot expected alpha net of modelled market impact against AUM and find where it crosses zero.
   2. Turnover is the multiplier. A signal with a two-day holding period has a fraction of the capacity of the same signal held for three months, because you pay the impact many more times.
   3. Crowding measures I would actually look at: short interest and days to cover, 13F overlap across similar funds, the share of float held by hedge funds, borrow fees trending up, and the beta of a name to a hedge fund crowding basket.
   4. The tell in the return series is a change in the character of the drawdowns. Crowded strategies lose money in sharp, correlated, liquidity-driven air pockets rather than in slow grinds, because everyone is selling the same thing on the same day.
   5. Two reference episodes make it concrete. August 2007, when quant equity books unwound together, and the early 2021 squeeze on crowded pod shorts. In both cases the fundamental signal was fine and the positioning was not.
   6. The honest limitation: crowding data is late. 13Fs are stale by 45 days, and by the time overlap is measurable the trade is already crowded. So I treat it as a sizing input and a reason to prefer less obvious expressions, not as a timing tool.

   Where candidates lose it

   Treating capacity as purely a signal decay story. The binding constraint is almost always market impact and turnover, and the crowding half of the answer needs specific observables: days to cover, 13F overlap, borrow trend. Vague talk about 'too much money chasing alpha' will not survive a follow-up.

   Expect next

   • How would you estimate capacity for a signal you just built?
   • What happened in August 2007?
   • How would you position differently if you knew a trade was crowded?
3. 033You are down 4 percent on the month and your risk manager is on the phone. What happens next?Risk and drawdownHardsuperdayMulti-manager platforms

   Say this

   First I would know the answer to their question before they ask it: what lost the money, whether it was the thesis or a factor, and what I am doing about it. Then I would cut risk to the level they need, without arguing, and keep the positions I still believe in at a smaller size.

   Then walk it

   1. Attribution first, and fast. Split the loss into market, sector, style factor and idiosyncratic. A 4 percent loss that is mostly a momentum unwind is a very different conversation from a 4 percent loss on two broken theses.
   2. If it is factor, the fix is mechanical: neutralise the offending exposure and the drawdown stops compounding. That is a risk process failure, and I would own it as such.
   3. If it is idiosyncratic and the theses are intact, I still have to reduce, because a drawdown limit is not an opinion. The question is which positions to keep. I would cut the ones where the falsifier has been triggered or where liquidity is worst, and keep the highest conviction, most liquid, nearest-catalyst names at reduced size.
   4. Say the number out loud, because it is how these seats work. Most platforms run a soft limit around 3 to 5 percent where gross is cut hard, and a hard stop near 7 to 10 percent where the book is closed. Knowing that is what shows you understand the seat.
   5. Then the behavioural discipline: do not double down to get it back, do not switch style, and do not stop communicating. PMs who go quiet in a drawdown get taken down faster than PMs who over-communicate.
   6. And the honest part: I would reserve the right to say that a position is being cut for risk reasons, not because I think it is wrong. That distinction is worth recording, because it is how you learn whether your process or your judgement failed.

   Where candidates lose it

   Saying you would defend the book and ask for more room. On a multi-manager platform the drawdown limit is the contract, not a negotiation, and that answer gets you marked as someone who will not survive the risk framework. Lead with attribution, accept the reduction, and show judgement in what you keep.

   Expect next

   • Which positions would you cut first, and why?
   • At what level does the platform close your book?
   • How would you tell the difference between bad luck and a broken process?
4. 034What is a drawdown, and why do funds care more about it than volatility?Risk and drawdownCoretechnicalMulti-manager platforms

   Say this

   A drawdown is the peak-to-trough fall in NAV, measured from the highest point reached. Funds care about it more than volatility because it is what triggers redemptions, stop-outs and the high water mark problem, all of which are path dependent in a way volatility is not.

   Then walk it

   1. The arithmetic is asymmetric and that is the whole point. Down 50 percent requires plus 100 percent to recover. Down 20 percent requires plus 25. Compounding punishes the depth of the hole, not the wiggle.
   2. Volatility is path independent and drawdown is not. Two funds with identical monthly volatility can have very different worst drawdowns depending on whether the bad months clustered.
   3. The commercial reason is redemptions. Investors leave near the trough, so a deep drawdown permanently shrinks the capital base and the manager never gets to earn the recovery on the original amount.
   4. Then the fee mechanics: below the high water mark the manager earns no performance fee until the loss is recovered, so a deep drawdown can make a business unviable even if the strategy eventually works. That is why funds sometimes close after a bad year rather than grind back.
   5. On a platform it is even more direct. The drawdown limit is a contractual stop, so a path that touches minus 8 percent and recovers is worse than a path that grinds to minus 5 and stays, because the first one ends your seat.
   6. Useful additional measures to name: time to recovery, the Calmar ratio which is return over maximum drawdown, and the Sortino ratio which only penalises downside deviation. Maximum drawdown alone is a single historical observation and therefore a fragile statistic.

   Where candidates lose it

   Defining drawdown correctly and then giving a purely statistical reason for caring. The reasons are commercial and structural: redemptions, the high water mark and the platform stop. Also do not present maximum drawdown as a robust risk measure. It is one realised path, and the next one will be different.

   Expect next

   • How long does it take to recover a 25 percent drawdown at a 10 percent return?
   • What is the Calmar ratio?
   • Why would a fund shut down rather than trade back to its high water mark?
5. 035Write a function that returns the n largest drawdowns in a return series.Risk and drawdownHardtechnicalBalyasny Asset ManagementQuantitative Trading · London · 2025

   Say this

   Build the cumulative NAV, walk it once tracking the running peak, and record a drawdown episode whenever the series falls below a peak and then makes a new high. Each episode gets a depth, a start, a trough and a recovery date. Then sort the episodes by depth and return the top n. It is a single linear pass.

   Then walk it

   1. Step one: turn returns into a wealth index, cumulative product of one plus r. Do this before anything else, because drawdowns are multiplicative and summing returns gives the wrong depth.
   2. Step two: running maximum of the wealth index. The drawdown series is wealth divided by running max, minus one, which is zero or negative at every point.
   3. Step three, the part interviewers actually test: segment into episodes. An episode opens when the drawdown series goes below zero and closes when it returns to zero, meaning a new high water mark. Within each episode the trough is the minimum.
   4. Step four: sort episodes by depth, take the first n. Say the complexity: O(T) for the pass plus O(k log k) for the sort, where k is the number of episodes, so linear in practice.
   5. State the edge cases before being asked, because this is where candidates get cut: the series ends while still in a drawdown, so the last episode is unrecovered and you should report it with no recovery date. Also decide whether overlapping nested dips count as one episode or several, and say which convention you are using.
   6. The naive alternative is to take the n most negative points of the drawdown series, and it is wrong: they will all sit inside the same crash. Volunteering why that fails is what shows you understood the question rather than pattern-matched it.

   Where candidates lose it

   Returning the n most negative values of the drawdown series. They cluster in one episode, so you report the same crash n times. The question is really about episode segmentation. Also, sum returns instead of compounding them and every number is wrong. State your episode convention out loud.

   Expect next

   • How would you handle a series that ends mid-drawdown?
   • How would you report time to recovery?
   • How would you do this for a portfolio of a thousand instruments efficiently?

   Reported by candidates at Balyasny Asset Management (Quantitative Trading, London, 2025). Source: Wall Street Oasis.

6. 036How does a stop-out work on a multi-manager platform?Risk and drawdownIntermediatetechnicalMulti-manager platforms

   Say this

   Each PM gets an allocation with a drawdown limit attached, usually a soft level where risk gets cut and a hard level where the book is liquidated and the seat ends. It is measured on the PM's own P&L against allocated capital, from the high water mark, and it is enforced by the risk team rather than negotiated.

   Then walk it

   1. Typical structure: a soft limit around 3 to 5 percent of allocated capital where gross is halved, and a hard stop around 7 to 10 percent where positions are flattened by the central desk.
   2. It is measured from the peak, so a PM up 6 percent has effectively earned a larger cushion for the year. That is why the first quarter of a new allocation is the most dangerous period in these seats.
   3. The mechanism is also a capital allocation tool. Risk budget is reallocated from PMs in drawdown to PMs performing, which is why the platform's own return stream is smoother than the average pod's.
   4. It is enforced centrally with real-time position and P&L monitoring, plus factor limits, single-name limits, liquidity limits and sometimes overnight and earnings-event restrictions.
   5. The behavioural effect is significant and worth being honest about. Hard stops truncate the left tail for the firm, but they also force selling into weakness, and they push PMs to run shorter horizons and narrower theses than they might otherwise. A structurally correct three-year view is not investable in a seat with a 6 percent annual stop.
   6. Second-order consequence: because platforms hire similar people and impose similar limits, a factor shock can force simultaneous deleveraging across many pods. The stop-out protects each fund and can amplify the market move.

   Where candidates lose it

   Describing it as a risk limit without the career consequence, or pretending you would simply never hit it. Say the numbers, say it is measured from the high water mark, and name the cost: hard stops force selling at the worst time and shorten every PM's investment horizon. Interviewers at platforms respect candidates who see both sides.

   Expect next

   • How would that change how you invest compared with a long-only seat?
   • What happens to your capital if you make money?
   • Is there a systemic cost to everyone having the same stop?
7. 037What is VaR, and what does it miss?Risk and drawdownIntermediatetechnicalRisk management

   Say this

   Value at Risk is the loss you would not expect to exceed on a given percentage of days. A one-day 99 percent VaR of 2 percent means that on 99 days out of 100 you lose less than 2 percent. What it misses is everything on the hundredth day, which is the day that matters.

   Then walk it

   1. Three ways to compute it: parametric from a covariance matrix, historical simulation from actual past returns, and Monte Carlo. Historical is the most common on an equity book because it makes no distributional assumption, but it can only show you crashes that already happened.
   2. The first big miss is the shape of the tail. VaR says nothing about how bad the bad day is. Expected shortfall, the average loss given you breached VaR, is the fix, and it is what Basel moved to for exactly this reason.
   3. The second is that it is not sub-additive in general. You can combine two books and get a VaR higher than the sum, which makes it a poor tool for allocating risk between pods.
   4. The third is estimation from a calm window. Volatility clusters, so a VaR calibrated on a quiet period understates risk precisely when positions have been built up on the basis of that calm.
   5. The fourth is what it cannot see: correlations going to one, liquidity vanishing, a borrow recall, gap risk over a weekend, and any non-linear payoff if you hold options and only use a linear approximation.
   6. So I would use it as a daily monitoring number and never as the risk decision. The decisions come from stress tests, liquidity analysis and position limits. As a one-line summary: VaR tells you how much you lose on a normal bad day, not on a bad day.

   Where candidates lose it

   Defining VaR and stopping, or getting the direction of the confidence level muddled. The question is explicitly about the limitations, so the second half of your answer is the answer. Name expected shortfall, name the calm-window calibration problem, and say what you would use instead.

   Expect next

   • What is expected shortfall and why is it better?
   • How would you stress test instead?
   • What is the danger of managing a book to a VaR limit?
8. 038Do you use stop losses on a fundamental book?Risk and drawdownIntermediatetechnicalLong-short equityMulti-manager platforms

   Say this

   Yes as a risk overlay, no as the investment decision. A price stop protects the fund from a thesis that is wrong in a way you have not yet understood. But on a fundamental book the primary sell trigger has to be the thesis breaking, and the two should be tracked separately.

   Then walk it

   1. The case for stops: a large adverse move is information. The market may know something you do not, and the discipline of cutting at a pre-set level protects you from the two biases that actually destroy books, anchoring and averaging down on conviction.
   2. The case against: fundamental theses need time, and a mechanical stop guarantees you sell your best ideas at the worst prices during factor rotations that have nothing to do with the company.
   3. So the workable version is an alert rather than an automatic sale. Down 15 percent triggers a mandatory re-underwrite from scratch, where I have to justify the position as a new purchase at today's price. If I cannot, it goes.
   4. Volatility-scaled levels are better than fixed percentages. A 15 percent stop on a 20 percent volatility utility and on a 70 percent volatility biotech are completely different statements about conviction.
   5. In a platform seat it is not a choice anyway. The drawdown limit at the book level effectively imposes stops at position level, and the honest framing is that the risk system owns the stop and the analyst owns the thesis.
   6. One number worth having: if your average winner is up 30 and your average loser is cut at 12, you can be right less than half the time and still make money. Say that and the question becomes about expectancy rather than about pride.

   Where candidates lose it

   Answering with a flat no because you are a fundamental investor, or a flat yes because it sounds disciplined. Both are one-sided. The credible answer separates the risk overlay from the investment decision and mentions the forced re-underwrite, which is what actually happens on good books.

   Expect next

   • Would you ever average down?
   • How would you set the level?
   • What is the difference between trimming and selling?
9. 039How do you think about transaction cost?Risk and drawdownIntermediatetechnicalMan GroupInvestment Management · Boston · 2022

   Say this

   In three buckets: the explicit costs, the spread, and market impact. The first two are small and easy to measure; impact is the big one and it scales with size, so for any strategy with turnover, cost is not a friction on the return, it is a constraint on the strategy.

   Then walk it

   1. Explicit: commissions, exchange fees, taxes such as stamp duty in the UK or STT in India. A few basis points, predictable, and in India securities transaction tax genuinely changes which strategies are viable.
   2. Spread: you cross half the bid-ask to trade immediately. In a liquid large cap that is one or two basis points; in a mid cap it can be 30.
   3. Impact: your own order moves the price. The standard working model is that impact grows roughly with the square root of the order size as a fraction of daily volume, so trading 10 percent of ADV costs far more than twice trading 5 percent.
   4. Then the cost nobody puts on the invoice: opportunity cost and delay. Trading slowly reduces impact and increases the risk the price runs away from you. That trade-off is exactly what an execution algorithm is solving, and implementation shortfall against the arrival price is the right way to measure the whole thing.
   5. The practical consequence is that cost has to be inside the signal, not after it. If a signal has 20 basis points of gross edge and 15 of round-trip cost, it should be traded slowly, netted against other signals, or not traded at all.
   6. One number to make it real: a book turning over 200 percent a year at 15 basis points round trip pays 60 basis points annually. That is the difference between a good year and an average one, and it is why netting flows across pods is a genuine advantage of a large platform.

   Where candidates lose it

   Answering with commissions and the spread and missing impact. Impact is the entire subject for anyone running real size, and the square-root rule plus implementation shortfall is the vocabulary that shows you have looked at execution data rather than a textbook. Also mention that cost belongs inside the signal, not subtracted afterwards.

   Expect next

   • How would you measure your own market impact?
   • How does cost change the optimal holding period?
   • What is implementation shortfall?

   Reported by candidates at Man Group (Investment Management, Boston, 2022). Source: Wall Street Oasis.

10. 040Your best long and your best short are in the same sector and both going against you. What do you do?Risk and drawdownHardsuperdayLong-short equityMulti-manager platforms

    Say this

    If both legs are losing at once, the pair is not hedged on the dimension that is moving, so first I would find out what that dimension is. It is usually a style factor rather than the sector, and the answer depends entirely on whether I have a factor problem or two independent thesis problems.

    Then walk it

    1. Diagnose before acting. Run the attribution: if the long is a value name and the short is a growth name, a growth rally hurts both and the sector hedge was never doing the work I thought it was.
    2. If it is a factor, the fix is to neutralise the factor, not to abandon the theses. Overlay a style hedge or adjust the pair weights so the loadings offset, and keep the idiosyncratic view.
    3. If both theses are genuinely deteriorating on their own facts, that is a different message: my process has a common flaw, probably a shared assumption about the end market. Then I cut both and go back to the work rather than pick one to defend.
    4. Check the correlation assumption I entered with. If I sized the pair as a hedged position and it is behaving as two directional positions, my risk is roughly double what I thought and the size has to come down immediately, regardless of the diagnosis.
    5. Then the liquidity ordering. Reduce where reducing is cheap, which often means cutting the more liquid leg first, and accept that this temporarily unbalances the pair.
    6. And I would say the part people skip: an adverse move in both legs of a pair is one of the most useful signals a book gives you, because it means your model of what the trade was exposed to was wrong. That is worth more than the P&L.

    Where candidates lose it

    Picking one leg to cut on instinct, usually the loser you like less. Without attribution you do not know whether you have one problem or two. Say what you would measure first. And if the pair was sized as a hedge but behaves directionally, the size is wrong now, so address that before discussing the theses.

    Expect next

    • How would you separate factor loss from idiosyncratic loss?
    • If it is a factor, do you hedge it or reduce?
    • What would you have done differently when you put the pair on?

Firm tags come from public, anonymous candidate reports on Wall Street Oasis: strong signal, not sworn testimony. Firms are named as the places a question was reported, not as partners of Fin Maverick. Answers are written for this page to show how to think out loud; they are not scripts to recite.