Risk Model Validation

1.A one-day VaR forecast is $3.2 million. The corresponding loss is $4.1 million. The test counts losses strictly greater than VaR. How should this observation be recorded?

2.A desk stores signed P&L and stores VaR as a positive loss threshold of $1.5 million. Which day's signed P&L is an exception under a strict exceedance rule?

3.A 97.5% one-day VaR forecast is tested over 400 valid days. Under correct unconditional coverage, what is the expected exception count?

4.A backtest has 8 exceptions across 200 valid observations. What is its observed exception rate?

5.Two 99% VaR backtests each have four exceptions over the same number of days. One has four consecutive exceptions, and the other has widely separated ones. Which additional property should be examined?

6.A one-day VaR report uses yesterday's closing positions. Today's actual loss includes an unusually large new trade made after the open. Which comparison helps isolate the frozen-book forecast?

7.Hypothetical loss is $900,000 and actual loss is $1.4 million against $1.1 million VaR. The reports use the same close and currency. Which result is accurate under separate strict tests?

8.A desk records a loss exception against hypothetical P&L, while actual P&L stays inside VaR after an intraday hedge. What should the investigation preserve?

9.An overnight VaR run uses a price file stamped later than the forecast cutoff. The newer file contains part of the move being forecast. What is the validation concern?

10.A $4 million loss exception appears on a day with a large external cash withdrawal. What should the analyst verify first?

11.A hypothetical backtest exception appears for a desk using close marks. One feed prices Europe at its local close and the hedge at a much later cutoff. What rerun tests the timing hypothesis?

12.A model shows one exception in 50 days at 99% confidence. What conclusion is supported by that observation alone?

13.A 99% VaR backtest produces zero exceptions in a short calm sample. What should a validator avoid claiming?

14.Six losses exceed VaR by $0.1 million each in one test. Another test has six exceptions that exceed it by $2 million each. Which statistic distinguishes them directly?

15.A desk backtests a ten-day forecast every day using overlapping ten-day P&L windows. What assumption needs care when testing exception independence?

16.A risk model's one-day VaR fell on Friday. The test compares it with a Monday-to-Wednesday cumulative loss. What is the first correction?

17.A portfolio appears to have 10 exceptions, but three use P&L from a different reporting currency than the forecast. What should happen?

18.A trader argues a single shock-driven exception should be deleted because the event was unprecedented. What is the sound response?

19.A backtest fails to produce an outcome on a day with missing prices. How should the validator report it?

20.A desk's exceptions begin the day a product mapping changes. What comparison best tests whether the change caused them?

21.An exception is dominated by one trade booked under a different legal entity than expected. Which evidence should be checked before blaming the model?

22.A frozen-book loss exceeds VaR on a day when one option's reference price appears to be ten times normal. What investigation is appropriate?

23.A risk-theoretical P&L series tracks hypothetical P&L except on large moves in credit spreads. What should the validator investigate?

24.A report finds a small average signed P&L difference between two pricing engines. The largest gaps occur precisely on loss days. What additional analysis matters?

25.A validator reproduces a prior exception using a position file downloaded today. The source can revise historical positions. What is missing?

26.A desk groups all exceptions across a large business. A small options book repeatedly fails but barely changes the group rate. What view is needed?

27.A model's average VaR rises after volatility jumps, but most exceptions occurred in the first week of the shock. What behavior does that suggest testing?

28.A validator compares a model before and after a fix. Which design avoids letting portfolio changes explain the result?

29.A one-day VaR exception results from an accounting reserve adjustment outside the documented risk forecast scope. What is the best write-up?

30.A backtest underpredicts losses mostly on days when a benchmark futures contract rolls. What is a targeted check?

31.A daily risk report shows a sharp reduction in VaR the day before an exception. What is the most useful attribution of the forecast change?

32.A risk model's exception count falls after switching from actual P&L to hypothetical P&L. What should the team report?

33.A challenger and production model both flag the same loss exception. The challenger forecast is much closer to the loss. Which comparison adds context?

34.A $2.8 million loss exceeds $2 million VaR. Repricing with a missing factor adds $0.6 million to the forecast, all else equal. What remains on the diagnostic rerun?

35.A model owner wants to label all disputed exceptions as “data” to close the review. What should validation require?

36.A reviewer has ten exception notes describing market events but none link to position sensitivities. What is missing?

37.A model applies normally distributed returns to a portfolio with repeated large downside jumps. Which challenge targets the assumption directly?

38.A historical simulation uses only the past six calm months. What should the validator test first?

39.A covariance model is calibrated to calm markets where equity and credit diversify each other. What scenario tests this assumption?

40.A model gives today and six months ago equal weight despite a persistent volatility regime change. What should a challenger compare?

41.A model estimates ten-day risk by multiplying one-day standard deviation by the square root of ten. Which assumption is central to that step?

42.A one-factor model prices all tenors from a single parallel yield shock. What is a direct test of its limitations?

43.A delta-only model measures an at-the-money option book. Which challenge is strongest for a large overnight shock?

44.A two-factor model reports zero net vega because front and back expiries offset. What should a challenger vary?

45.A model marks thin corporate bonds using a liquid sector index. Which evidence tests whether the proxy is fit for purpose?

46.A data vendor fills missing daily marks with the previous price. What should validation measure?

47.A challenger fits volatility parameters after looking at the entire backtest period, then compares itself with forecasts that were generated live. What makes the comparison fair?

48.A vendor model is proprietary and has limited documentation. What can validation still test?

49.Two pricing engines agree on liquid instruments but diverge on callable bonds. What is a targeted comparison?

50.A new model treats foreign bonds as if prices and FX rates moved independently. What joint shock can challenge the assumption?

51.A model simulates stress but assumes every asset can be liquidated at its shocked midpoint. What sensitivity should validation add?

52.A historical model discards every return beyond three standard deviations as a bad tick. What should be required?

53.A risk engine estimates vega by bumping volatility. A 10-point bump gives a materially different per-point sensitivity from a 0.1-point bump. What should validation distinguish?

54.A model's input documentation calls a volatility shock “2,” but implementation treats it as 2.00 in decimal volatility. The intended shock was two percentage points. What should the input be?

55.A risk model assigns an interest-rate-sensitive preferred stock to an equity index alone. Which test targets the missing mapping?

56.A portfolio has many long positions on related companies. A model uses diversification estimated from an index with few concentrated names. What should the validator compare?

57.A volatility estimate appears stable because intraday data is sampled only once a day. Which review tests fitness for an intraday limit?

58.A benchmark model and production model use the same vendor marks and pricing library. They agree closely. What limitation remains?

59.A model performs well for linear futures and is proposed for a desk with path-dependent options. What should be tested before approval?

60.A model scenario bumps all credit spreads by the same amount. What test challenges its treatment of distressed issuers?

61.A scenario assumes implied volatility and spot move independently. A portfolio sells downside puts. Which challenge is most relevant?

62.A stress model reports a $1.2 million loss under its baseline correlation. Under a plausible stressed correlation, loss is $2.0 million. What does the comparison show?

63.A model assumes costs are proportional to trade size. The desk's largest positions exceed normal market depth. What evidence challenges linearity?

64.A factor history starts after a new instrument launches and covers only quiet markets. Which use of a longer proxy history is defensible?

65.A model selects the shortest calibration window that minimizes exceptions on a known crisis. What is the key challenge?

66.A report uses annualized volatility in one leg and daily volatility in another. What should be fixed before calculating covariance-based risk?

67.A model's assumed mapping from a spread factor to an instrument has a beta of one. What controlled test checks the implementation?

68.A challenger improves coverage by doubling every forecast, including during stable markets. What else should validation compare?

69.A developer changes a model's sampling weights and adds a new factor in the same release. What test isolates their effects?

70.A model calibration is refreshed daily, but parameters are clipped at hard upper bounds that were set years ago. What should be tested?

71.A new model uses an industry average recovery for all defaulted issuers. Which challenger tests whether this assumption is material?

72.A validation lead must decide whether a simpler model is acceptable for an approved limited use. Which evidence is most persuasive?

73.A dashboard shows total DV01 near zero. Two-year DV01 is +$8,000 and ten-year DV01 is −$8,000, where positive gains when yields fall. Two-year yields rise 10 bp and ten-year yields fall 5 bp. What loss does the first-order curve test show?

74.A spread portfolio has $5,000 of long spread DV01 per bp. Its benchmark rate hedge leaves rate DV01 near zero. A 40 bp spread widening occurs at unchanged rates. What loss does the first-order test show?

75.A model reports $900,000 stress loss before funding costs. A multi-day replay adds $150,000 of financing costs and $220,000 of liquidation costs that are absent from the original. What is the expanded loss estimate if no effects overlap?

76.A short-option book has near-zero delta and negative gamma. A committee sees a small delta estimate for an overnight gap. What should the analyst show?

77.A report nets long and short foreign exposures across different currencies by dollar amount. What example exposes the blind spot?

78.A hedge reduces variance in ordinary markets but fails whenever trading in the underlying is halted. What limitation belongs in the risk report?

79.A credit report uses net exposure, but two positions are with different counterparties and cannot be netted for settlement. What additional view reveals risk?

80.A model understates exposure because it applies a 100-share multiplier to a contract that delivers 150 shares. It reports +600 shares of delta. With the same per-share delta, what is corrected exposure?

81.A daily dashboard omits working orders. It shows 80 shares long under a limit of 100, with two executable buy orders for 15 each. What exposure should a worst-case fills check report?

82.A fund's market risk forecast uses closing positions, but a strategy routinely carries larger intraday exposures that it flattens by close. What should the reviewer explain?

83.A risk dashboard shows no increase after a private holding misses three scheduled valuations. What warning and check are appropriate?

84.A production model predicts $1.1 million loss. A challenger predicts $1.9 million. Adding the same missing FX basis factor to production raises its estimate by $0.5 million. What difference is still unexplained?

85.A model groups all options into a single parallel vega number. Portfolio vega is +$300 per point overall, but downside-wing vega is −$900. Under a wing-only rise of 3 points, what is the first-order result?

86.A model's baseline assumes two equal risk sleeves have zero correlation and each has standalone volatility of $1 million. What is joint volatility if correlation rises to +1, compared with roughly $1.41 million at zero correlation?

87.A purchased credit hedge pays only when a named issuer defaults, but the long portfolio contains other issuers in the same sector. What test highlights a model that treats the hedge as a perfect offset?

88.A model measures cash-bond exposure using only clean prices but the loss report contains carry, coupon receipts, and accrued interest. What should an explanation do?

89.A risk report shows lower gross exposure after a change in how derivatives are measured, but the positions are unchanged. What should validation check?

90.A scenario estimates liquidation at 10% of normal volume. In stress, total daily volume is 40% of normal and the maximum participation remains 10% of available volume. How does liquidation time change for a fixed position?

91.A model reports $3 million loss and the approved temporary overlay adds $800,000 for missing liquidity costs. What figure and explanation should the limit report give?

92.A temporary overlay was introduced after missing downside skew caused exceptions. What evidence should support removing it?

93.A validator finds that a new bond type is absent from model coverage. The desk still wants to trade it. What interim proposal is reviewable?

94.A committee is told a model “misses options risk.” Which statement makes the gap actionable?

95.A challenger predicts larger losses on every date, but it also includes a broader portfolio than production. What comparison supports a model-specific conclusion?

96.A risk report estimates $2 million loss under a moderate shock and $5 million under a severe one. What wording avoids overstating certainty?

97.A validator proposes tightening a limit while a defect is repaired. What should the risk owner see?

98.A pricing library is upgraded, and model results shrink. The vendor reports higher accuracy overall. What should the analyst examine before accepting lower exposure?

99.A risk model measures market moves but excludes variation margin cash flows. A desk uses the number to decide whether it can survive two days of stress. What should be added?

100.A portfolio's hedge tracks a benchmark well before transaction costs. A stressed replay shows the hedge can only be adjusted in small clips. What is the most accurate explanation?

101.A validation run finds that a stressed correlation change raises forecast loss by $700,000 only on one concentrated book. Which statement should management receive?

102.A model's option delta estimate agrees with live P&L for small moves. A full reprice loses $2.4 million in a gap while the delta estimate loses $0.8 million. What should the report say?

103.A model reports less risk after several positions are reclassified into a diversified group. The holdings and market data did not change. What is a targeted review?

104.A firm has corrected a unit error in a material risk factor. What post-fix check most directly shows the correction remains in place?

105.A model runs correctly but a desk uses it outside its validated horizon to approve large positions. What is the main problem?

106.A portfolio manager asks whether a validation finding means the system's number should be ignored entirely. What is the most useful response?

107.A committee wants a short explanation of repeated breaches during volatility spikes. The model uses slow volatility updates and an illiquid hedge proxy. Which recap is strongest?

108.A desk reports $1.4 million of modeled loss, $2.2 million under an independent repricing, and $2.8 million after adding realistic exit costs to that repricing. Which closing note best supports a decision?