Part A of this glossary covered the foundational vocabulary and the risk metric terms — sensitivity, delta, vega, curvature, VaR and Expected Shortfall. FRTB Part B finishes the job: how the standard talks about hedging and diversification, how it decides whether a risk factor can be modelled at all, the vocabulary used to validate internal models, and the two terms specific to CVA risk.
A quick note on scope: several of the terms in this part — modellability, backtesting, P&L attribution — belong primarily to the Internal Models Approach, which sits outside this series’ core focus on the Standardised Approach. We are still covering them in full here, both because MAR10 defines them as part of the same glossary, and because Article 1 established that even IMA banks must calculate the Standardised Approach in parallel — so understanding where the two approaches’ vocabulary overlaps and where it diverges is genuinely useful context.
Terminology for Hedging and Diversification
| Term | Plain-English Meaning |
| Basis risk | The risk that two instruments in a hedging strategy do not move in perfect lockstep with each other, which weakens how well the hedge actually protects the position. |
| Diversification | The reduction in overall portfolio risk that comes from holding risk positions across different instruments whose values are not perfectly correlated with one another. |
| Hedge | The process of offsetting risk by holding both long and short risk positions in instruments that are correlated with each other. |
| Offset | Narrower than a hedge: specifically, netting long and short risk positions that sit in the exact same risk factor. |
| Standalone | Being capitalised on a standalone basis means a set of risk positions is booked in its own separate, non-diversifiable trading book portfolio — it cannot diversify, hedge, or offset against risk elsewhere, and nothing else can diversify, hedge, or offset against it either. |
| Where You’ve Already Seen “Standalone” in This Series This exact concept appeared back in Article 1: when a bank uses the Internal Models Approach, it must still calculate the Standardised Approach for every trading desk, treating each desk as a standalone portfolio with no cross-desk offsetting. Now you have the formal definition behind that requirement — no diversification, hedging, or offset benefit is allowed to cross the standalone boundary in either direction. |
Sourced from MAR10.21–MAR10.25. https://www.bis.org/basel_framework/chapter/MAR/10.htm
Terminology for Risk Factor Eligibility and Modellability
| Term | Plain-English Meaning |
| Real prices | A price counts as “real,” for the purpose of testing whether a risk factor is eligible, if it comes from one of three sources: an actual transaction the bank itself conducted, an actual transaction between other arm’s-length parties (such as on an exchange), or a firm quote — a price at which the bank could actually have transacted with an arm’s-length party. |
| Modellable risk factor | A risk factor that qualifies as modellable based on how many representative real price observations exist for it, plus additional qualitative principles about the data used to calibrate the Expected Shortfall model. A risk factor that fails this eligibility test is instead classified as a non-modellable risk factor (NMRF). |
The modellable-versus-non-modellable distinction is one of the more consequential ideas in the entire FRTB framework, even though it sits almost entirely on the Internal Models Approach side. Risk factors with too few observable real prices cannot be modelled reliably, so they get pulled out of the standard Expected Shortfall calculation and capitalised separately as NMRFs — typically at a more punitive capital charge, reflecting the market’s own lack of confidence in pricing that risk.
Sourced from MAR10.26–MAR10.27. https://www.bis.org/basel_framework/chapter/MAR/10.htm
Terminology for Internal Model Validation
| Term | Plain-English Meaning |
| Backtesting | Comparing a bank’s actual and hypothetical daily profits and losses against what its VaR model predicted, to check how conservative — or how accurate — its risk measurement really is. |
| Profit and loss (P&L) attribution (PLA) | A check on how robust a bank’s risk management models are, done by comparing the risk-theoretical P&L predicted by a trading desk’s risk model against the desk’s hypothetical P&L. |
| Trading desk risk management model | For an in-scope desk, this includes every risk factor in the bank’s Expected Shortfall model that carries supervisory parameters, plus any risk factor judged non-modellable — those NMRFs are excluded from the ES capital calculation itself, but are still part of this broader risk management model and get capitalised through the separate NMRF charge. |
| Actual P&L (APL) | The real daily P&L from the bank’s actual P&L process — including intraday trading, time effects, and new or modified deals — but excluding fees, commissions, and valuation adjustments that already have their own separate regulatory capital treatment or are deducted from Common Equity Tier 1. Other market-risk-related valuation adjustments must be included. APL also includes FX and commodity risk from banking book positions. |
| Hypothetical P&L (HPL) | The P&L you would get by revaluing yesterday’s closing positions using today’s market data. Commissions, fees, intraday trading, new or modified deals, and the same excluded valuation adjustments as APL are left out, though valuation adjustments that are updated daily are usually included. Time effects must be treated consistently between the HPL and the risk-theoretical P&L below, so the two stay comparable. |
| Risk-theoretical P&L (RTPL) | The daily desk-level P&L predicted by the valuation engines inside the trading desk risk management model, using every risk factor in that model — including the NMRFs. |
| Why Three Different Versions of “Daily P&L” Exist APL, HPL and RTPL all describe a bank’s daily profit or loss, but each is built to answer a different question. APL is what genuinely happened, including all the messiness of intraday trading and new deals. HPL strips that messiness out, asking purely “what would yesterday’s book be worth using today’s prices?” — which makes it comparable to a VaR model’s own prediction. RTPL goes a step further, asking what the desk’s own risk model would have predicted for that same move, including the NMRFs the ES capital calculation itself excludes. Comparing these three versions against each other is exactly what backtesting and P&L attribution are for. |
Sourced from MAR10.28–MAR10.33.
Terminology for Credit Valuation Adjustment Risk
| Term | Plain-English Meaning |
| Credit valuation adjustment (CVA) | An adjustment made to the valuation of a derivative transaction to account for the credit risk of the parties on the other side of the contract. |
| CVA risk | The risk that CVA itself changes — driven by changes in the credit spreads of the contracting parties, and made worse or better by changes in the value, or the variability of the value, of whatever the derivative is written on. |
These two terms connect directly back to Part D of our internal risk transfers mini-series, where we covered how internal risk transfers between the CVA portfolio and the trading book are treated, and the specific rules preventing the same risk from being capitalised twice under both the CVA framework and the market risk framework.
Sourced from MAR10.34–MAR10.35. https://www.bis.org/basel_framework/chapter/MAR/10.htm
The Complete MAR10 Glossary at a Glance
| Category | Terms Covered | Where |
| General terminology | Market risk, notional value, trading desk, pricing model | Part A |
| Financial instruments | Financial instrument, instrument, embedded derivative, look-through approach | Part A |
| Capital requirement calculations | Risk factor, risk position, risk bucket, risk class | Part A |
| Risk metrics | Sensitivity, delta, vega, curvature, VaR, Expected Shortfall, JTD, liquidity horizon | Part A |
| Hedging and diversification | Basis risk, diversification, hedge, offset, standalone | Part B |
| Risk factor eligibility | Real prices, modellable risk factor / NMRF | Part B |
| Model validation | Backtesting, P&L attribution, trading desk risk management model, APL, HPL, RTPL | Part B |
| CVA risk | CVA, CVA risk | Part B |
Frequently Asked Questions
What is a non-modellable risk factor (NMRF)?
A risk factor that fails the eligibility test for modellability — typically because it lacks enough representative real price observations — and is therefore excluded from the standard Expected Shortfall calculation and capitalised separately instead.
What is the difference between Actual P&L, Hypothetical P&L and Risk-Theoretical P&L?
Actual P&L is what genuinely happened on the desk, including intraday trading and new deals. Hypothetical P&L revalues yesterday’s positions with today’s market data, stripping out that daily activity. Risk-Theoretical P&L is what the desk’s own risk model would have predicted for the same move, including non-modellable risk factors.
What is the difference between a hedge and an offset?
A hedge counterbalances risk across long and short positions in correlated instruments. An offset is narrower — it specifically nets long and short positions within the same risk factor.
https://decode-finance.com/category/market-risk-credit-risk-and-operational-risks