How are you modeling event risk (geopolitics/war) inside a quant futures/options framework?
I'm an analyst on a multi-commodity book and I've been building out models to take a more systematic view on futures and options across the complex. The core feature set I'm working with right now:
- Market structure / term structure (curve shape, carry, roll yield)
- Cross-commodity correlation and spreads
- Seasonality
- Fundamentals (S&D, flows, inventories)
That part feels reasonably well-trodden. Where I keep getting stuck is event risk, geopolitics, conflict, sudden policy/export shocks. These are the moves that actually drive the tails, but they're hard to fit into a feature set built on continuous, mean-reverting-ish variables.
A few things I'd genuinely like to hear other people's experience on:
- How do you represent a discrete shock (war, export ban, sanctions) in a model that's otherwise fundamentals/structure driven? Regime dummy, jump component, separate overlay, or do you just exclude those periods?
- Does anyone actually model these as binary/digital payoffs — i.e. P(event) x payoff — rather than trying to forecast the price path? Curious whether that holds up out of sample or just looks clean on paper.
- On the feature side: what's earned its place in your models and what looked great in backtest but added nothing live?
Happy to trade some notes.
And happy to network if anyone here is working in Switzerland.