Mock Screens
Risk Analyst Screen
A single 25-minute, 18-item sitting modelled on bank and clearinghouse risk screens - VaR, Expected Shortfall, correlation breakdown, fat tails, and short gamma, with no negative marking.
What is a bank/clearinghouse-style risk analyst screen?
Candidates for market-risk, counterparty-risk, or clearing-risk analyst roles at banks and clearinghouses are typically given a timed, no-calculator screen before any interview, because the job runs on reproducing a small set of standard formulas - VaR, Expected Shortfall, portfolio variance - correctly and fast, under exactly the kind of pressure a real trading day produces.
These screens favor multiple choice and short numeric answers over open-ended case work, because the skill being tested is not creativity but reliability. A risk desk cannot tolerate an analyst who understands VaR in theory but fumbles the arithmetic when a limit is breached mid-session, so the format rewards candidates who have the core formulas memorized cold rather than derived from first principles each time.
A large share of the questions are not computational at all - they test whether a candidate understands why a metric exists and where it breaks down: why Expected Shortfall replaced VaR in bank capital rules, why correlations converge to 1 in a crisis, why fat tails make a normal-based model dangerous. Firms ask these because a risk analyst's real job is flagging exactly the situations where the standard model is quietly wrong, not just plugging numbers into it.
Because clearinghouses and bank risk desks operate under regulatory scrutiny - Basel-style capital rules, margin methodologies - screens for these roles also probe whether a candidate can reason about model risk itself: what a metric fails to capture, and what happens when its assumptions (normality, stable correlation) are violated. A representative sitting mixes "compute this" items with "what does this consequence mean" items in the same short list, rather than testing arithmetic alone.
How it works
The sitting is a single section, "Tail risk & portfolio stress" - 18 items in 25 minutes, drawn from one pool: parametric VaR and Expected Shortfall calculations, why Basel moved from VaR to ES, correlation-breakdown and diversification-collapse reasoning, fat-tail consequences, and short-gamma behavior under a volatility spike. The standard normal values the calculations need (z0.99 = 2.3263, the ES multiplier 2.6652) are given in the instructions, so nothing has to be looked up.
Items are served one at a time, either numeric-entry or four-option multiple choice, and the clock runs continuously for the whole 25 minutes with no per-item limit. allowBack is false: once an item is submitted or left blank it cannot be revisited and the answer-sheet palette cannot jump backward. Flagging an item does not let you return to it during the run, but it does put the item - with the correct answer and its explanation - on the marked script afterward.
The section ends automatically when the clock reaches zero or the 18th item is used, whichever comes first. Anything left over is simply scored as skipped, with no further penalty.
How scoring works
Each of the 18 items is worth 1 mark. A correct answer scores +1; the section's penalty is 0, so a wrong answer scores 0, not negative - the code computes each section's net as correct minus wrong times the penalty, which with penalty 0 reduces to the raw count correct. An unanswered item also scores 0.
Because wrong and blank score identically, there is no scoring reason to leave an item blank instead of guessing - the "no negative marking" rule in the instructions is exact. The review screen reports the raw score out of 18, the percentage of maximum, and a per-skill accuracy breakdown (distributions, expected-value, options-greeks) built only from items you actually answered.
No percentile is reported - the review screen states this directly, since there is no comparison population behind this assessment yet.
Tail risk & portfolio stress
Roughly half the pool is pure arithmetic - VaR = z x sigma x NAV, or ES = sigma x 2.6652 x NAV - and the instructions hand you both multipliers, so the only work is multiplying three numbers and rounding to two decimals. Do the arithmetic in percentage units (multiply the vol and NAV first, then apply the multiplier) rather than converting to dollars at every step; it is one fewer place to drop a decimal under a running clock.
The wrong choices on every VaR/ES item are built from the same three numbers with one formula element swapped - z0.99 replaced with z0.95, the z-multiplier dropped entirely, or the ES multiplier used where VaR was asked for. Reading the stem for which of VaR or ES is being asked, and at which confidence level, matters more than the multiplication itself, because the distractors are built to reward a candidate who computed correctly but read carelessly.
The conceptual items - why Basel moved to ES, what happens to portfolio vol when correlations go to 1, what fat tails do to a normal-based model, what short gamma does under a volatility spike - have one defensible answer each and no calculation. If you have not internalized why ES is subadditive and VaR is not, or why correlation convergence removes diversification exactly when a portfolio needs it most, those items cost more time than any arithmetic item.
Pacing across the whole sitting
25 minutes across 18 items is about 83 seconds each on average, but items are not evenly hard: budget 15-20 seconds for a conceptual item you know cold, and spend the time saved there on the VaR/ES arithmetic.
Because allowBack is false, flag anything you are unsure of rather than laboring over it in place - a flagged item's explanation still shows up on the marked script afterward, while time spent stalled on one item is time the clock is taking from the rest of the 18.
With no negative marking, submit your best answer on every item before the clock reaches zero. A rushed final answer scores the same as a careful one, and strictly better than a blank.
A worked example
The item gives a book size and a daily volatility, and asks for the 99% one-day parametric VaR assuming zero mean.
z0.99 = 2.3263 is given directly in the instructions, so no lookup is needed. Here sigma = 1.5% and NAV = $200M.
Substitute the volatility as a decimal fraction and multiply through.
Round to two decimals - this is the number the item checks against. Enter $6.98M.
The three wrong choices swap one piece of the formula each: z0.95 instead of z0.99, no z-multiplier at all, and the ES multiplier used where VaR was asked for. Recognizing the pattern is the real skill.
Common mistakes
• Confusing VaR and Expected Shortfall - the stem says which one it wants, and the ES multiplier (2.6652) is always larger than z0.99 (2.3263), so mixing them up produces a wrong answer that still looks plausible.
• Recomputing the full correlation-weighted variance formula for the rho = 1 item instead of using the shortcut - once every pairwise correlation is 1, portfolio volatility is just the weighted sum of the individual volatilities, no matrix arithmetic needed.
• Leaving an item blank instead of guessing. With penalty 0, blank and wrong both score zero - there is never a reason to skip an item you have any read on.
• Reasoning that a leptokurtic (fat-tailed) return series makes a normal-based VaR model overstate risk. It is the opposite - fat tails put more probability mass in the extreme, so a normal model systematically understates tail losses.
• Trying to go back to fix an earlier answer. allowBack is false for this section - once you submit or skip, the item is gone until the marked script at the end.
Why interviews test this
Every bank or clearinghouse risk-analyst interview eventually asks you to compute a VaR or ES figure from a volatility and a notional, and to explain why regulators moved capital requirements from VaR toward Expected Shortfall. This format tests exactly that pairing - arithmetic plus the conceptual reason the arithmetic isn't the whole story. A candidate who can produce the number but not explain why ES is preferred (subadditivity, and that it describes the whole tail rather than pointing at its edge) reads as someone who memorized a formula without understanding what it protects against.
The correlation-breakdown and short-gamma items map to the follow-up question interviewers ask after the VaR calculation: what happens to this number in a crisis? Being able to say, without hesitation, that diversification is a property of the correlation structure and disappears exactly when correlations converge to 1 - and that a short-gamma position's losses accelerate rather than stay linear as a move grows - is what separates a candidate who has only studied the formulas from one who has thought about when they fail.