pybaw 1.0.0

Barone-Adesi-Whaley (1987) American option pricing and implied volatility (Cython)

pybaw

Barone-Adesi–Whaley (1987) American option pricing and implied volatility as a Cython extension: nogil scalar loops with scipy's Cython brentq doing the root solves.

Conventions

• sigma is the Black total vol (total standard deviation, sigma_annual * sqrt(T)).
• The pricer works in (D, F, S, sigma) terms — rT = -ln(D), bT = ln(F/S) — so no explicit T appears (T=1 convention).
• Dividends never appear explicitly: the cost of carry is implied via carry = F/S.
• baw_implied_vol takes the raw premium (not undiscounted); its intrinsic is the American max(S-K, 0) / max(K-S, 0). Sentinels: 0.0 at/below intrinsic, NaN on failure.

Install

pip install pybaw

Prebuilt wheels cover CPython 3.11+ on Linux (x86_64, aarch64), Windows (x86_64), and macOS (arm64, x86_64). On anything else pip falls back to the sdist, which needs a C compiler — the compiled extension is required, there is no pure-Python fallback. Note scipy is needed at build time too, not just runtime: the extension cimports scipy.optimize.cython_optimize.

To work on the package itself, either install from a checkout:

pip install .

or build in place and put the repo root on PYTHONPATH:

python setup.py build_ext --inplace

API

Function	Purpose
baw_price(S, K, D, F, sigma, flag)	American price; flag 'c'/'p'
baw_implied_vol(price, S, K, D, F, flag)	invert a raw premium to total vol
vectorized_baw_implied_vol(prices, S[], K[], D[], F[], flags[])	batch inversion
find_forward_baw(calls, puts, strikes, spot, discount)	forward where call IV ≈ put IV near ATM
bsm_price(K, D, F, sigma, flag)	pure European reference (used by tests/bounds)

find_forward_baw exists because European put-call parity is biased for American options by the early-exercise premium; solving for the forward that equates call and put implied vols near the money removes that bias.

Tests

pip install -e .[test]
pytest

License

MIT — see LICENSE.