Where Models End and Underwriting Begins

Every catastrophe model converges on a single number: expected loss.

Decades of data. Geospatial hazard maps. Probability theory – all compressed into a single output. But that output is only as good as its assumptions. And assumptions are not facts.

Every catastrophe model relies on:

• Historical loss distributions
• Return period estimates (1-in-50 year, 1-in-100 year, 1-in-250 year events)
• Exposure growth curves
• Climate trend adjustments

Each input is a best estimate, not a fact. And that distinction matters more today than ever.

The Problem With “Stationarity”

Traditional catastrophe modelling assumes something called stationarity – the idea that the statistical properties of extreme events don’t materially change over time.

In simple terms:

If a 1-in-100-year flood happened historically once per century, the model assumes that pattern broadly holds. But climate science is increasingly clear: stationarity is breaking down.

We are seeing:

• Shifting hurricane intensities
• Changing precipitation patterns
• Urbanization amplifying loss severity
• Secondary perils (flood, wildfire, convective storm) driving more volatility than primary perils

The data updates, but reality often moves faster than the data.

That creates a structural gap.

The Model-Reality Gap

Catastrophe models are backward-looking by necessity. They learn from historical distributions and adjust incrementally. But underwriting lives in the present. That creates a fundamental tension in reinsurance.

Models provide a baseline. Underwriters price the uncertainty around that baseline.

Two reinsurers can run the exact same model and arrive at very different prices – not because the math is different, but because the judgment is.

The real question isn’t

"What does the model say?"

It’s

"How wrong might the model be?"

Why Underwriting is Still an Art

This is why underwriting has never been purely quantitative.

The best reinsurers in the world, from Lloyd’s syndicates to the Bermuda Class of 2001 to modern catastrophe bond desks, have always treated pricing as a dialogue between quant and judgment.

The model outputs an expected loss. The underwriter decides:

• How much margin to demand
• Where climate trends are underpriced
• Which cedents manage exposure best
• When capital is mispriced in the cycle

That layer – the judgment layer – is where alpha has historically lived in reinsurance. It’s also the least visible part of the system.

The Transparency Problem

Traditional reinsurance is built on trust.

You allocate capital to a syndicate, a reinsurer, or a fund, and you rely on track record, reputation, and sparse disclosures. You rarely see how capital is deployed, how underwriting decisions evolve, or how risk appetite shifts mid-cycle.

In most cases, the underwriting “art” is a black box. You either trust it –  or you don’t.

Bringing Underwriting Onchain

At OnRe, we’ve taken a different approach. We don’t pretend underwriting can be fully automated. You cannot remove judgment from risk transfer – and you shouldn’t try. What you can do is make that judgment observable.

That’s the philosophy behind ONyc. The yield generated through ONyc reflects real underwriting decisions applied to real risk transfer in global reinsurance markets. But unlike traditional structures – where allocators receive quarterly reports – we pair that with transparency primitives native to crypto:

• Independent attestations
• Active capital disclosures
• Onchain utilization metrics
• Verifiable deployment data

Instead of asking for blind trust, we expose the inputs.

Transparency Without Oversimplification

There’s a temptation in crypto to believe everything should be purely algorithmic. In markets like lending or AMMs, that works. In reinsurance, it doesn’t.

Reinsurance operates in non-stationary environments with sparse data and structural tail risk. You cannot compress that into a single formula. The goal isn’t to eliminate the art. It’s to make the science, and its application, visible.

A New Model for Risk Markets

Traditional reinsurance asked allocators to trust opaque underwriting judgment. Purely algorithmic DeFi tries to remove judgment entirely. We believe the future sits between those extremes. A model where:

• Quantitative science remains foundational
• Underwriting judgment remains essential
• Transparency becomes programmable

Where capital providers can interrogate not just the outputs, but the process.

Where Art Meets Science

Risk modelling will never be perfectly certain – not in a world where climate, exposure, and capital flows are constantly evolving. But markets don't need perfect certainty, they need credible frameworks and aligned incentives.

At OnRe, that means giving allocators something traditional reinsurance rarely has: visibility into the decisions that drive their returns.

Because the real edge in risk markets has always lived in that space:

Where models end – and underwriting begins.