Computational studies

These notebooks develop curve construction, valuation and risk calculations using the Python binding. Each study introduces a financial problem, states its assumptions and interprets the calculated results. Read the Python guide for the conventions shared by the studies, and use the reference when a declaration needs clarification. The pages include the code, tables and figures saved in each canonical notebook.

The pages render saved notebook outputs without executing Python during a documentation build. To reproduce the saved results, use the companion inputs and a binding built from the same checkout. The notebooks record their numerical assumptions, including grids, random seeds and comparison tolerances.

Curve construction

Construct a discount curve from market quotes and examine its forwards, zero rates and repricing checks.

• From deposit and OIS quotes to a discount curve

Swaps and collateral currencies

These studies progress from a receive-fixed position to par-rate solving and then to cashflows, FX direction, and collateral-dependent currency conversion.

• A receive-fixed swap: cashflows, value and risk
• A par swap coupon from discounted cashflows
• Valuing a CLP swap under USD collateral

Fixed income and loans

Bond quotation anchors, issuer calls, contractual amortization, and borrower prepayment are different economic problems. The studies keep their price and cashflow conventions separate.

• Price, yield, duration and Z-spread for a fixed-rate bond
• A callable stepped-coupon bond
• Understanding mortgage cashflows and value

Rate options and volatility

Inspect option inputs before interpreting a price: a rate strike, a quoted Black volatility, and a calibrated short-rate sigma are not interchangeable.

• Valuing a caplet and a cap
• Rate and volatility drivers of option value

Short-rate simulation

The simulation study examines calibration, stored paths, seeds, and sampling uncertainty. Materialized simulation paths do not by themselves carry calibration or volatility sensitivities.

• Calibrated short-rate paths and their distribution

Portfolio risk and benchmarking

Start with portfolio membership, position sizes, requests and aggregate result tables. Then inspect weighted coupons, fixing indices and separate interest/capital payments, and reconcile sensitivities with full revaluation. The legacy benchmark is an optional external comparison, not an additional pricing API: its ORE/CurveEngine prerequisites are separate from QuantSupport and are not required to build this documentation or run the main tutorial sequence. Its reported agreements for curves, PV, and rate risk do not establish a validated FX hedge: fully propagated FX spot risk differs from the historical JSON’s -V/S rescaling derivative. The study reports that disagreement explicitly rather than hiding it with a comparison tolerance.

• How a portfolio works
• Portfolio aggregation: coupons, payments and risk
• Quote risk and revaluation for a mixed portfolio
• Comparing curves, swap values and quote risk across engines

Scripted payoffs

The scripting study compares compatible native and scripted representations. Distinguish deterministic checks from finite-path estimators and from control-adjusted selected result variables.

• Dated payments: from a caplet to an amortizing note

Exposure and XVA

Compare netting and collateral assumptions, signed exposure means, and CVA/FVA. An exposure profile is neither a potential-future-exposure quantile nor a time-averaged exposure measure.

• Netting-set exposure and XVA

Reproduction

From the repository root, install the example dependencies into the environment containing the freshly built native package, then run:

python -m bindings.python.examples.run_notebooks
python -m bindings.python.examples.run_notebooks valuation/swaps --write

The runner starts a clean kernel for each selected study and uses that interpreter unless a kernelspec is explicitly selected. The first command validates without modifying source notebooks; --write refreshes stored outputs. Before running cells it probes the actual kernel and binding exports, rejecting stale bindings. Saved execution_environment metadata records sanitized Python/package versions and whether the kernel matched its caller; inspect this provenance rather than inferring identity from a launch command. The legacy external benchmark lives outside that runner’s directory but is available through its benchmark selector. To validate all sixteen, install the benchmark’s optional requirements and include it explicitly:

python -m pip install "open-source-risk-engine==1.8.14.1" "curveengine @ git+https://github.com/jmelo11/curveengine.git@d7e000da626e08ff2be5d34561597065340b017e"
python -m bindings.python.examples.run_notebooks --include-benchmark

Every rendered study links its canonical source and companion-file directory so a downloaded notebook is not mistaken for a self-contained market-data bundle.