How the model works
Crisis Lighthouse is built on a deliberately simple idea: each asset class carries a set of fixed sensitivity weights describing how it has historically tended to respond to six types of shock. Your slider settings combine with those weights to produce an outlook score. Everything is on this page — there is no black box.
The scoring formula
Each slider runs from 0 (shock not happening) to 100 (extreme severity). Each asset holds a weight from −3 (historically hurt badly) to +3 (historically benefits strongly) for every scenario. The raw score is simply the sum of weight × severity across all six scenarios.
That raw sum is then passed through a tanh squashing function and scaled to a −100 to +100 display range. The squash matters when you stack several crises at once: instead of every asset pinning at the extremes, scores approach ±100 asymptotically, so even in a compound catastrophe you can still see which assets are more exposed than others.
The "drivers" shown under each asset name are the scenarios currently contributing most to its score, so you can always trace a number back to its cause.
The complete sensitivity matrix
These are the exact weights used by the dashboard. Positive numbers are tailwinds in that scenario; negative numbers are headwinds.
| Asset | War | Recession | Disaster | Pandemic | Inflation | Energy |
|---|---|---|---|---|---|---|
| Defence & aerospace | +3 | −0.5 | 0 | −0.5 | 0 | 0 |
| Energy producers (oil & gas) | +2 | −2 | +0.5 | −2 | +2 | +3 |
| Healthcare & pharma | 0 | +1 | +0.5 | +3 | 0 | 0 |
| Consumer staples | +0.5 | +1.5 | +1 | +2 | +0.5 | −0.5 |
| Technology (growth) | −1 | −1.5 | 0 | +2 | −2 | −1 |
| Cybersecurity | +2.5 | −0.5 | 0 | +1.5 | −0.5 | 0 |
| Financials & banks | −1 | −2.5 | −1 | −1.5 | +0.5 | −0.5 |
| Consumer discretionary & luxury | −2 | −3 | −1 | −1.5 | −1.5 | −1 |
| Airlines, travel & tourism | −3 | −2 | −2 | −3 | −1 | −3 |
| Utilities | 0 | +1.5 | −1 | +1 | −1 | +0.5 |
| Industrials & construction | +0.5 | −2 | +2 | −1 | −0.5 | −1.5 |
| Agriculture & food producers | +1.5 | +0.5 | +2 | +1 | +1.5 | +0.5 |
| Gold | +3 | +1.5 | +1 | +1.5 | +2.5 | +0.5 |
| Crude oil & natural gas | +2.5 | −2.5 | +1 | −2.5 | +2 | +3 |
| Industrial metals (copper, silver) | +0.5 | −2 | +1 | −1 | +1.5 | 0 |
| Agricultural commodities (grains) | +2 | 0 | +2.5 | +0.5 | +1.5 | +1 |
| Residential property | −1.5 | −1.5 | −2 | +0.5 | +1 | −0.5 |
| Commercial real estate | −1.5 | −2.5 | −2 | −3 | +0.5 | −1 |
| Government bonds (Treasuries) | +1.5 | +2.5 | +0.5 | +2 | −3 | −1 |
| Inflation-linked bonds (TIPS) | +1 | +0.5 | +0.5 | +0.5 | +3 | +1.5 |
| Corporate high-yield bonds | −1.5 | −2.5 | −1 | −2 | −1.5 | −0.5 |
| Cash & short-term deposits | +1 | +1.5 | +1 | +1 | −2.5 | −0.5 |
| Safe-haven currencies (USD, CHF, JPY) | +2 | +1.5 | +0.5 | +1.5 | −1 | 0 |
| Cryptocurrency | −0.5 | −2 | 0 | +1 | +0.5 | −1 |
Where the weights come from
The weights encode broad tendencies documented across historical crisis episodes — the ones discussed in detail in our scenario guides: 1973 and 1979 for energy shocks, the 1970s and 2022 for inflation, 2008 for recession and credit crunch, 2020 for pandemic, Katrina and Fukushima for disasters, and conflicts from the Gulf War to Ukraine 2022. They are judgment calls informed by history, not statistical estimates — reasonable people could argue any weight up or down half a point, and we consider that a feature: everything is visible and debatable.
What the model deliberately ignores
- Valuations. An asset that is cheap going into a crisis behaves differently from the same asset entering expensive. The model has no opinion on today's prices.
- Policy response. The 2020 crash became a boom because of unprecedented stimulus. The model cannot know what central banks and governments will do.
- Timing and sequencing. In real crises, everything falls together in the liquidity-panic phase before winners separate from losers. Scores describe the episode, not each week of it.
- Second-round surprises. Every real crisis contains something no matrix anticipated. That is what makes it a crisis.
Use it as a thinking tool. The dashboard is a structured way to reason about scenarios and stress-test a portfolio's logic — not a prediction engine, and not financial advice.