Context
The diagnostics report mathematical evidence (min density, Lee slopes, calendar margins). Missing is the economic layer: a tiny violation in an untradeable wing is not an executable arbitrage, and a mathematically clean surface can sit far outside every bid/ask. SVI also happily evaluates at K=1e-9 with no signal that this is pure extrapolation.
Proposal
- Classify findings: mathematical arbitrage vs quote-consistent (violation within bid/ask uncertainty) vs executable (constructible static portfolio from quoted, crossed prices) vs extrapolation risk (violation outside the quoted domain)
- Domain-of-validity metadata on evaluation: every slice records its quoted range;
surface.iv(..., return_status=True) labels results observed / interpolated / extrapolated
- Reports show the classification alongside the existing numerical evidence
Acceptance criteria
- Wing-violation fixture outside the quoted range classified as extrapolation risk, not executable arbitrage
- Crossed-quotes fixture classified as executable
- Evaluation status labels covered by tests
Context
The diagnostics report mathematical evidence (min density, Lee slopes, calendar margins). Missing is the economic layer: a tiny violation in an untradeable wing is not an executable arbitrage, and a mathematically clean surface can sit far outside every bid/ask. SVI also happily evaluates at K=1e-9 with no signal that this is pure extrapolation.
Proposal
surface.iv(..., return_status=True)labels results observed / interpolated / extrapolatedAcceptance criteria