A dependency-light Python ecosystem for general actuarial work — from experience analysis and projection through rating, reserving, and loss modeling to tail estimation and portfolio capital.
Small, composable packages that take and return plain pandas and numpy, share one
set of conventions — tidy tables, explicit distribution parameterizations,
reproducible randomness — and build on a common primitives layer, actuarialpy,
rather than a framework you have to buy into.
| Package | Role |
|---|---|
| actuarialpy | Calculation primitives, the canonical Experience data contract, and the ExperienceSet workbook layer (one construction call over source tables) |
| experiencestudies | Experience reporting, actual-vs-expected, claimant and concentration analysis |
| projectionmodels | Claim, premium, and expense projection over a renewal horizon |
| ratingmodels | Manual and experience rating, credibility, indication, GLM relativities |
| reservingmodels | Claims development and stochastic reserving: chain ladder, Bornhuetter-Ferguson, Mack, and the ODP bootstrap of the predictive reserve distribution; builds triangles from a claims-listing Experience |
| lossmodels | Severity and frequency fitting, aggregate loss distributions; fits directly from a claims-listing Experience |
| extremeloss | Extreme-value tails: POT/GPD, GEV, return levels, splicing; fits directly from a claims-listing Experience |
| risksim | Portfolio Monte Carlo, dependence, reinsurance contracts, risk measures |
pip install openactuarial # all eight packages, one tested set
pip install actuarialpy # - or any package on its ownopenactuarial is a meta-package with no code of its own: it pins the exact
versions of all eight packages that were tested together as a release train, so
a single command installs a known-good combination, and each release bumps those
pins to the next tested train. Prefer looser ranges? Install the packages
individually — each declares its own compatible range (>=X.Y,<X.Y+1). Requires
Python 3.10+.
Full API reference and worked examples that run end to end: openactuarial.org. Every package's test suite reruns nightly against the current PyPI releases (the ecosystem smoke), so cross-package drift surfaces within a day.
MIT — see LICENSE.