Scratio

Risk Analytics

Estimate portfolio risk under simulated or historical market scenarios.

Overview

Risk Analytics provides methods for estimating potential portfolio losses under different market scenarios.

The Quant Risk API analyzes portfolio risk by generating or selecting possible market outcomes, calculating the corresponding portfolio values, and measuring the resulting profit and loss distribution.

Different risk models use different approaches to produce portfolio outcomes. Some models generate simulated market scenarios from statistical distributions, while others use historical market observations as scenarios.

The resulting portfolio outcomes are used to calculate risk metrics such as Value at Risk (VaR) and Conditional Value at Risk (CVaR).

How Risk Estimation Works

Although the underlying methods differ, the risk estimation process follows a common high-level structure.

Define the portfolio and the market data used by the selected risk model.
Generate or select market scenarios.
Calculate the corresponding portfolio outcomes for each scenario.
Construct a distribution of portfolio profit and loss outcomes.
Calculate risk metrics from the resulting distribution.

The specific calculations performed at each stage depend on the selected risk model.

Risk Models

The API provides multiple approaches for estimating portfolio risk. Each model differs primarily in how market scenarios and portfolio outcomes are produced.

Monte Carlo Normal

Monte Carlo Normal generates correlated asset return scenarios using a multivariate normal model estimated from historical return data.

Monte Carlo Student-t

Monte Carlo Student-t generates simulated return scenarios using a Student-t distribution. This approach can model heavier tails than a normal distribution.

Historical Simulation

Historical Simulation estimates portfolio outcomes by applying historical market return scenarios rather than generating new scenarios from a statistical distribution.

Portfolio Outcomes

Each market scenario produces a possible portfolio outcome over the selected time horizon.

The portfolio outcome is typically represented by the difference between the portfolio value at the end of the scenario and its initial value.

\operatorname{PnL} = V_{\mathrm{terminal}} - V_{\mathrm{initial}}

$\operatorname{PnL}$ represents the portfolio profit or loss for an outcome.
$V_{\mathrm{terminal}}$ represents the portfolio value at the end of the selected time horizon.
$V_{\mathrm{initial}}$ represents the initial portfolio value.

Positive P&L values represent gains, while negative P&L values represent losses.

The collection of outcomes across all scenarios forms a portfolio P&L distribution that can be used to measure risk.

Risk Metrics

Risk metrics summarize information from the distribution of portfolio outcomes.

Value at Risk

Value at Risk estimates a loss threshold at a selected confidence level over the specified time horizon.

VaR focuses on the boundary of the loss distribution at the requested confidence level.

Conditional Value at Risk

Conditional Value at Risk measures the average loss among outcomes in the selected tail of the distribution.

CVaR provides additional information about the severity of losses beyond the VaR threshold.

The exact calculation of these metrics may depend on the methodology and implementation used by the selected risk model.

Choosing a Method

The appropriate risk model depends on the assumptions that are suitable for the portfolio and the type of market behavior being analyzed.

Monte Carlo Normal may be suitable when asset returns are modeled using a multivariate normal distribution.
Monte Carlo Student-t may be useful when modeling return distributions with heavier tails.
Historical Simulation may be appropriate when historical market movements are used directly as the source of risk scenarios.

Each method involves different assumptions and may produce different estimates of portfolio risk from the same underlying portfolio data.