Decision guide¶
Use this page to route into the strongest engineering proof in the repo. The proof sequence has three steps: repair a noisy implied-vol surface, make the Dupire handoff smooth enough to trust, then show local-vol and PDE repricing evidence.
Choose by question¶
| If you want to show... | Open this page | Supporting notebook | Why it matters |
|---|---|---|---|
| The library can repair and diagnose a noisy quote surface | Surface repair workflow | 06_surface_noarb_svi_repair.ipynb | This is the clearest static-surface engineering proof: quotes, diagnostics, SVI fit quality, and repair are all visible. |
| The local-vol handoff is smoother than a slice stack | eSSVI smooth handoff | 07_essvi_smooth_surface_for_dupire.ipynb | This is where the repo proves the time-derivative problem is understood and explicitly improved. |
| The numerics are validated rather than just implemented | Local-vol and PDE validation | 08_localvol_pde_repricing.ipynb | This page shows repricing accuracy, error structure, and convergence rather than a black-box PDE result. |
| The benchmark story is real and reproducible | Performance evidence | Benchmark artifacts under benchmarks/artifacts/ |
This is where scaling, remedy tradeoffs, and end-to-end stage budgets are measured. |
| The system design is deliberate and typed | Architecture | 09_surface_to_localvol_pde_integration.ipynb | The architecture page explains how the pricing, volatility, numerics, and validation layers fit together. |
Proof sequence¶
- Read Surface repair workflow.
- Continue to eSSVI smooth handoff.
- Finish with Local-vol and PDE validation.
Use Performance evidence and Architecture after that sequence when the follow-up questions turn to scaling, cost, or system design.