reinsurance.dev

Find a topic

Foundations

The industry

• The market
• Market participants
• Products and contracts
• Perspectives on loss
• Personas

Analytics Toolkit

Trials and loss tables

• What is a trial?
• Events and occurrences
• The Year Event Loss Table (YELT)
• Event Loss Tables

Analytics Toolkit

Curves and metrics

• Exceedance probability (EP) curves
• Occurrence (OEP) vs. aggregate (AEP)
• Expected loss (EL)
• Value at Risk (VaR)
• Tail Value at Risk (TVaR)
• Attachment and exhaustion probability

Financial Modelling

Terms

• Loss filter
• Contract period
• Occurrence excess
• Aggregate excess
• Scaling

Financial Modelling

Contracts

• Quota share
• Catastrophe excess of loss (CatXoL)
• Aggregate excess of loss (AggXoL)
• Dissecting a contract
• Contracts as compositions

Financial Modelling

Programs and portfolios

• Financial perspectives
• Programs
• Portfolios

Financial Modelling

Beyond loss

• Flows that are not loss
• Reinstatement premium
• Two reinstatements, and a clause pro rata to time
• Pricing with reinstatement premium

Applications

Business decisions

• Risk profiling: the post-model report
• Standalone program pricing
• Portfolio analytics
• At production scale

Reference

Caldera Re

• Company profile
• The portfolio: six contracts
• Data files

Reference

Vocabulary and conventions

• Glossary
• Scope, conventions & assumptions
• Sign convention

Who this is for

This site is written for software engineers who are new to reinsurance: the backend engineer who has just joined a reinsurer with strong software skills and no insurance background, the senior engineer pivoting from fintech, the architect modernising a legacy risk system. It assumes you can read Python and reason about probability. It assumes nothing about insurance. Actuaries will find the mathematics stated precisely; engineers will find every formula that computes a number paired with the code that computes it.

Scope

Property-catastrophe reinsurance, after the model has run. Everything on this site starts from the catastrophe model’s loss output (thousands of simulated futures, each a trial in the vocabulary the Analytics Toolkit defines) and works downstream: how contracts transform loss, how portfolios aggregate it, and how prices and capital derive from it. The focus is deliberate. Property cat is the line where a single event can cost the industry tens of billions.¹ This site chose it because the loss data is rich enough, and the contracts intricate enough, to make the analytics an engineering problem in their own right.

Not covered, by design

The following topics are outside this site’s scope. Each is a discipline of its own with its own methods and literature. They sit either upstream of the cat model’s loss output or outside property catastrophe:

• Catastrophe-model internals: hazard, vulnerability, and event-generation modules. This site consumes the model’s loss output; it does not explain how the model produces it.
• Exposure rating and exposure curves: pricing from exposure profiles rather than model output.
• Experience rating: pricing from a cedent’s historical losses.
• Reserving and IBNR (incurred but not reported): estimating liabilities for losses that have already occurred.
• Casualty, specialty, life, and health reinsurance: lines whose analytics rest on different foundations than simulated catastrophe output.

If a question sits upstream of the model’s loss output, or outside property cat, the question is out of scope here.

The scope statement defines the subjects this site covers. Other conventions (vocabulary, the sign convention, units, loss perspectives, capital) are indexed on one page, Scope, conventions & assumptions, each with a pointer to the section that explains it.

How the site is organised

Four chapters, in reading order, and three reference pages the chapters draw on throughout:

A single fictional reinsurer, Caldera Re, appears in every chapter: six contracts, five cedents, and a 20-trial loss dataset small enough to check by hand. The numbers recur from page to page, so a figure in the Toolkit is the same loss that a contract transforms in Financial Modelling and a pricing story prices in Applications.

Footnotes

1. Insurance Information Institute, “Facts + Statistics: Global catastrophes”, citing Aon data as of January 2026. Its table of the ten costliest world natural disasters by insured loss, 1900 to 2025, is headed by Hurricane Katrina at $65 billion insured ($107 billion in 2025 dollars) and Hurricane Ian at $54 billion. ↩