Risk Measurement and Control: Trading Guide

Before entering the second ring — If risk per trade, position size, notional value, leverage, margin, liquidation, drawdown and risk of ruin are not yet distinct, begin with Risk management in trading. This chapter asks a broader question: not only “how much is this position risking?” but “how do we measure and control vulnerabilities across the whole system?”.

A risk measure is not risk itself. It is a constructed lens with a scope, horizon, dataset, assumptions and aggregation rule. VaR and Expected Shortfall inspect an estimated distribution; stress tests and scenarios ask what would happen under stated shocks; aggregation and diversification depend on how positions relate; liquidity, operations and intermediaries can make realised outcomes depart from the model.

This is a chapter on established risk practice. It neither introduces the Emiciclo Method nor turns prudential standards written for banks into universal rules for every trader.

Interactive map of the second ring

Explore the thirteen nodes. The map separates three tasks: measuring outcomes, combining exposures and keeping the process operational.

Lens	The question it answers	The mistake to avoid
Value at Risk	Which estimated loss quantile corresponds to the stated level and horizon?	calling it “maximum loss”
Expected Shortfall	What is estimated average loss in the tail beyond that quantile?	treating it as the worst possible case
Stress testing	What does an explicit severe shock do to the chosen perimeter?	mistaking the result for a forecast
Scenario analysis	How does a coherent narrative propagate through factors and positions?	assigning precision the scenario does not have
Model risk	Where can data, assumptions, implementation or use produce adverse decisions?	limiting control to the formula
Aggregate risk	What is risk across the selected scope, and what contributes to it?	adding incompatible measures
Correlation	How strong is the observed linear relation between two series?	inferring causality, stability or tail behaviour
Diversification	How do weights, factors and dependencies distribute risk?	counting instruments instead of finding concentrations
Counterparty risk	What loss may follow if the other party fails to perform?	always identifying it with the broker or venue
Liquidity risk	At what time, cost and impact can a position be built, reduced or exited?	equating observed volume with guaranteed liquidity
Funding liquidity	How will payments, margin and collateral needs be met when due?	inferring it from order-book liquidity alone
Operational risk	Which failures of people, processes, systems or third parties can disrupt activity?	reducing resilience to a static checklist
Broker risk	What follows from the entity, contract, custody and execution arrangements?	treating a licence as universal protection

1. Distribution: VaR marks a threshold; ES looks beyond it

For a defined loss variable L, VaR at a chosen level is a quantile of its estimated distribution. The number needs coordinates: unit or currency, scope, horizon, confidence level, valuation date and method. “VaR 100,000” is not interpretable without them.

Expected Shortfall instead considers the average loss in the tail beyond the quantile, with technical care for discontinuous distributions. It therefore adds information about the severity of tail outcomes that the single VaR cut-off does not describe. ES is still not the largest possible loss, and it remains sensitive to data, valuation and model choices. The Basel market-risk framework defines ES as the average of potential losses exceeding VaR at a given level; the framework's particular regulatory parameters belong to that application and are not defaults for every portfolio.

Backtesting, exception analysis, alternative benchmarks and model governance help locate discrepancies between the lens and experience. They do not make a future distribution true, and they do not replace stress tests.

Model risk runs through the entire process. Incomplete data, unsuitable assumptions, implementation defects or use outside the intended scope may lead to adverse decisions even when the arithmetic is correct. Inventory, documentation, independent challenge, monitoring and management of limitations are distinct controls; their intensity should reflect materiality and the consequences of model use.

2. Stress and scenarios: make vulnerability explicit

A stress test applies defined adverse conditions and translates their effects into value, P&L, capital, collateral, liquidity or another relevant resource. The family includes several methods:

sensitivity analysis changes one or a small set of factors;
a historical scenario reuses observed movements against the current book;
a hypothetical scenario combines coherent shocks that need not have appeared together in the dataset;
a reverse stress test starts from a result that would make the system unviable and searches for combinations capable of producing it.

Quality is not the most dramatic number. It comes from the objective, coverage of material risks, severity, consistency across factors, non-linearity, second-round effects, liquidity, documentation and use of results. BCBS places sensitivity analysis, complex scenarios and reverse stress inside the broad stress-testing family and calls for proportionate application; it does not prescribe one universal calendar for every user.

3. Portfolio: aggregate without concealing

For a linear portfolio viewed through variance, the compact expression σ²p = wᵀΣw shows why weights and covariances act together. It is not a universal formula for risk. Non-linear payoffs, asymmetric losses, defaults, funding, concentrations and liquidity constraints require additional views.

Correlation standardises covariance and describes a linear relationship in a sample. It depends on the series, sampling frequency, window and regime; it does not prove causation or promise unchanged dependence in stress. Therefore diversification is a measured effect, not a count of tickers. Different funds may load on the same factors, a hedge may retain basis risk, and small net exposure may coexist with large gross exposure, leverage or liquidation risk.

A legible aggregation preserves at least:

a consistent perimeter, currency and timestamp;
gross and net exposure without conflating them;
concentrations by position, factor, sector, currency, venue and counterparty;
contributions to the chosen measure;
dependence assumptions in current and stressed conditions;
reconciliation to positions and data-quality controls.

Counterparty risk adds current and potential exposure, default, recovery, collateral, netting and wrong-way risk. The counterparty must be identified by function and contract: broker, venue, custodian and clearing house may be different entities, and an intermediary acting as agent is not automatically the economic counterparty to the trade.

4. Where the model meets the world: liquidity, operations, brokers

Market liquidity concerns whether the intended quantity can be traded with acceptable time, cost and price impact; spreads, depth, immediacy and resiliency observe different dimensions. Funding liquidity is the capacity to meet payments, margin and collateral calls when due. They may amplify each other: urgent sales move prices, those moves trigger further margin needs, and the available response time contracts.

This is why funding liquidity deserves a separate lens: maturity schedules, available cash, usable collateral, concentrated sources and a contingency funding plan cannot be read from the order book alone.

Operational risk covers events arising from people, processes, systems or external events. Operational resilience adds the outcome of continuing and recovering critical operations through disruption. Prevention, detection, containment, continuity, reconciliation and learning therefore belong to one control cycle.

A broker is at once a legal entity, a contractual relationship, an execution channel, a possible custodian and a technology dependency. Due diligence starts with identity and jurisdiction, then covers authorisation, disciplinary history, ownership and segregation of assets, applicable protections, conflicts, execution quality, withdrawals and continuity. Regimes have different limits: in the United States, for example, BrokerCheck exposes registration and disclosure information, while SIPC explains that its protection does not cover market loss and does not extend indiscriminately to every instrument.

Two reading routes

For a newer reader — Value at Risk → Expected Shortfall → stress testing → model risk → diversification → liquidity risk → broker risk. The goal is to understand why no single number closes the problem.

For portfolio work — scenario analysis → aggregate risk → correlation → diversification → counterparty risk → funding liquidity → operational risk. The goal is traceability across data, dependencies, contributions, limits and responses.

The third ring: allocating and reading exposures

After aggregate measures, the next step is to separate capital, risk budgets and contributions, then read beta, option Greeks and rate sensitivities in their respective units. The third ring connects measurement to the actual structure of exposures without compressing everything into one number.

Open Risk allocation and sensitivities →

Sources

This chapter is general education. It does not estimate the risk of a specific portfolio, replace contractual or regulatory documentation, or make an investment recommendation.