MODEL SELECTION WORKBENCH
Pick the first model before writing the simulator.
Turn your surface, data, and compute constraints into a ranked research path. Scores are planning heuristics, never prices or forecasts.
Prototype first
CGMY
Best fit for asymmetric jumps, with a heavier simulation and calibration burden.
What changed
The sample favors jump flexibility over implementation speed.
Weakest assumption
The score assumes dense enough strikes to identify four CGMY parameters.
Do next
Freeze one date, calibrate CGMY and Merton, then compare residual smile structure and runtime.
Trade-offs
See the reason, not just the rank.
Each score is a weighted planning signal from the controls above.
| Model | Jump realism | Vol dynamics | Path cost | Calibration | Prototype |
|---|
Transparent scoring assumptions
What this cannot tell you
These scores do not estimate option values, forecast returns, test arbitrage constraints, or replace calibration against clean market data. They only order a research backlog from stated preferences.
Durable output
Leave with a testable decision.
Save the scenario, restore it later, or export the assumptions and next experiment for review.
Representative sample loaded.