Brazilian VIX

A VIX-style forward-looking volatility index for the Ibovespa, computed from B3 option chains. B3 publishes no VIX equivalent, so the index is built from scratch and validated against realised volatility.

Constraint

The CBOE VIX methodology assumes inputs that B3's listed option market does not supply cleanly. Adapting it meant choosing, at each step, between theoretical fidelity and producing a signal at all from the strikes actually quoted. The code favours the latter and states where that trade-off was made.

Data

Every run fetches over HTTP and holds results in memory. There is no database.

Source	Used for	Code
OpLab REST API (api.oplab.com.br/v3)	Option chains, interest rates	oplab_api.py — fetch_options_data(), fetch_interest(), get_historical_options()
BRAPI (brapi.dev/api)	Prime rate	BrAPIWrapper.py
yfinance (^BVSP)	Ibovespa closes, for realised volatility	VolCalculation.ipynb

Requires OPLAB_API_KEY in the environment.

Method

1. Pull the IBOV option chain and the prevailing interest rate.
2. Select strikes and expiries adapted to what is actually quoted.
3. Compute a daily forward-looking volatility level.
4. Compute realised volatility from Ibovespa closes and compare.

Result

Over the sample in VolCalculation.ipynb, the index and realised volatility have a Pearson correlation of 0.7531 — the index tracks the direction of realised volatility.

This is a co-movement result, not a calibration result. The two series are not on a comparable scale: the index ranges roughly 1194–3189 while realised volatility ranges 5.22–15.05, and the comparison cell applies a hand-tuned × 50 rescale purely to overlay them on one chart. The correlation is invariant to that rescale, so the 0.7531 stands on its own — but the index is not yet expressed in units that can be read as a volatility percentage, and no RMSE, R² or regression fit is computed. Levels should not be compared across the two series until the scaling is derived rather than fitted by eye.

Known issues

• Scaling is unresolved. The index is not in interpretable volatility units (see above). The × 50 factor sits under a comment that says 100 — neither figure is derived.
• oplab_api.py:675 is broken. It does from Brapi import BrAPIWrapper; the module is BrAPIWrapper.py, so calculate_vix_df raises ModuleNotFoundError on that path.
• The notebook was executed out of order across sessions, so stored execution counts are not sequential.

Running

pip install -r requirements.txt
export OPLAB_API_KEY=...
jupyter notebook VolCalculation.ipynb

License

MIT. See LICENSE.

---

Pedro Todescan