Ulcer Index Constraints

PortfolioOptimisers.set_risk_constraints! — Method
set_risk_constraints!(
    model::Model,
    i,
    r::UlcerIndex,
    opt::Union{AbstractProgrammeAllocationSet, RiskJuMPOptimisationEstimator},
    pr::AbstractPriorResult,
    args...;
    prefix,
    kwargs...
) -> Any

Add Ulcer Index risk constraints to model.

Introduces a scalar variable uci and one second-order cone over the drawdowns, then defines the risk expression uci_risk. Without observation weights, the cone is [sc * uci; sc * dd[2:T+1]] in SecondOrderCone() and uci_risk = uci / sqrt(T). With observation weights w, each drawdown is scaled by sqrt(w_t) inside the cone and uci_risk = uci / sqrt(sum(w)). The keys are indexed by i, because two Ulcer Index measures with different weights are two expressions.

Mathematical definition

\[\begin{align} \mathrm{UCI}(\boldsymbol{w}) &= \frac{\lVert \boldsymbol{dd} \rVert_2}{\sqrt{T}} = \sqrt{\frac{1}{T}\sum_{t=1}^T dd_t^2}\,. \end{align}\]

Where:

  • $\mathrm{UCI}(\boldsymbol{w})$: Ulcer index.
  • $\boldsymbol{dd}$: Drawdown vector.
  • $T$: Number of observations.
  • $dd_t$: Portfolio drawdown at time $t$.

With observation weights, the weighted mean of the squared drawdowns is used instead:

\[\begin{align} \mathrm{UCI}(\boldsymbol{w}) &= \sqrt{\frac{1}{W_{T}}\sum_{t=1}^T w_{t} dd_t^2}\,. \end{align}\]

Where:

  • $w_{t}$: Observation weight of observation $t$.
  • $W_{T} = \sum_{t=1}^{T} w_{t}$: Total observation weight.

Arguments

  • model::JuMP.Model: The JuMP optimisation model.
  • i: Constraint index for unique variable and constraint naming.
  • r::UlcerIndex: Ulcer index risk measure instance.
  • opt::RiskConstraintOwner: The owner of the risk constraint, a risk-based JuMP optimisation estimator or a programme Allocation Set.
  • pr::AbstractPriorResult: Prior result containing the returns matrix X.

Validation

Returns

  • uci_risk: The Ulcer Index risk expression.

Related

source