qfwp 0.4.2

Companion package for 'Learning Quantitative Finance with Python' (CRC Press): real historical episodes (qfwp.data) and the Trone synthetic market (qfwp.synth).

qfwp — companion package

Companion code for Learning Quantitative Finance with Python (CRC Press).

uv add qfwp      # or: uv sync, to build .venv from pyproject.toml + uv.lock

from qfwp import synth, data

mkt = synth.make_market(level=4)          # the full Trone market
mkt.prices.plot()                          # equities + TRN/USD
data.load("devaluation_1999")              # fetches BCB PTAX, caches locally

Two subpackages:

• qfwp.data — the real historical episodes (1994 Real Plan, 1999 devaluation, 2008 GFC, 2011 EUR/SNB, 2014 elections). Live: loaders fetch on demand from official open-data services (BCB SGS, FRED, ECB) and cache to $QFWP_CACHE (default ~/.cache/qfwp); set QFWP_OFFLINE=1 to pin a build to the cache. No data is redistributed — each series rides under its source's terms, not the package's MIT licence. See SOURCES.md.

• qfwp.synth — the Trone synthetic market with a four-level complexity dial matching the book's four parts, all four implemented:

  • 1 (Basics) — one asset, Trone calendar, seven injectable data pathologies;
  • 2 (Derivatives) — rates curve + meetings, FX pair, vol surface with skew, option chains on the meeting cycle;
  • 3 (Statistics) — six-asset fat-tailed panel, calm/stress regime, t-copula tail dependence (correlations rise in stress);
  • 4 (Risk) — credit names with wrong-way spreads and a jump-to-default, liquidity (widening spreads, no-quote days), and the central-bank floor that holds for years then breaks.

  Levels are strict supersets and the dial design is in SYNTH-DESIGN.md.

notebooks/discrete_hedging.ipynb reproduces the Derman–Kamal discrete-hedging result on the Trone market (see Chapter 8).

Everything is seeded and pinned (uv.lock), so the book's figures reproduce exactly — same seed, same numbers, same chart, on any machine.