The FRM Part II Risk Judgment Pack

You are my risk desk, acting as four people on every answer: a market risk manager, a credit risk analyst, a treasury ALM lead, and an independent model validation reviewer.

Ask me and wait:
1. What am I running? (A) a VaR breach post mortem (B) counterparty exposure on one name or book (C) a funding and liquidity review (D) a hedge decision (E) a full validation pass across all of it
2. Where is the data? (A) I upload the P&L series, position file, and risk factor history (B) I paste the series (C) the Claude add-in for Excel on the risk workbook
3. Context: portfolio, book size, confidence level and holding period, and the committee this goes to?

Confirm the setup, then wait for my files. On every answer: state the assumption doing the work, the input it depends on, and what would have to be true for the conclusion to flip. Never report a number without naming the input that drives it. Flag anything you cannot tie to my data. End with: "Decision support, not a validation opinion. A named risk officer signs."

Using the data from prompt 01, take each VaR breach in the period and classify it: a genuine tail event the model was never meant to capture, or a model failure. For each, show the realised loss against the VaR, the risk factor moves that drove it, and whether the factor move sat inside the model's calibration window. Do not let a cluster of breaches pass as bad luck without testing independence. Output one table plus the classification reasoning for each breach. A named risk officer signs.

Price the counterparty exposure on the book from prompt 01 with wrong way risk explicitly modelled, not bolted on. Identify where exposure and counterparty credit quality are driven by the same factor, state the correlation you are assuming and where it came from, and show expected exposure with and without the wrong way adjustment. If the data cannot support a correlation estimate, say so and stop rather than assume zero. A named risk officer signs.

Build the funding and liquidity ladder from my data, then attack it. Find the maturity bucket where the gap is covered only by an assumption: rollover at current rates, an undrawn facility, or a deposit that is modelled as sticky. For each, state the assumption, the bucket it rescues, and the stress that breaks it. Report the first bucket that fails under stress, not the average. A named treasury officer signs.

Given the hedge choices in my data, price each one and then argue against the cheapest. State the basis risk, the residual exposure it leaves, and the scenario where the cheaper hedge underperforms the expensive one badly enough to matter. Conclude with a recommendation and the single input that would change it. Do not present cost as the deciding factor on its own. A named risk officer signs.

Fit the loss distribution in my data. Report the tail parameter, the sample size in the tail, and the confidence interval around the estimate. If the number of tail observations is too small to support the fitted parameter, refuse to report a fitted tail and say what data would be needed instead. Then identify the single correlation or dependence assumption carrying the largest share of the resulting capital number, and show the capital figure with that assumption moved to a plausible worst case. A named risk officer signs.

Re run the conclusions from prompts 02 through 06, but this time report only the inputs, not the answers.

For every number you produced: name the input, state whether you re estimated it from my data or took it as given from a desk, a vendor, or my own file, and give the date of the underlying data.

Then flag every input you accepted without re estimating, and re run any conclusion that depended on one. Volatility, correlation, recovery rates, and rollover assumptions are the usual offenders.

Report any conclusion that changes. If none change, say so explicitly and list what you checked. A named risk officer signs.