We watch a set of geopolitical situations that move markets, score how far along each one is, and connect them to companies that would actually feel it. Everything below is the plain-English version. You do not need any of it to use the Desk.
A specific situation that would move prices if it escalated — a shipping strait becoming dangerous, a border exchange turning into something larger, a supply chain losing a chokepoint. We track a few dozen of them.
Each one gets a short code (T8, T2 and so on) so we can tag a company to it and you can see, at a glance, which situations a stock is exposed to.
How far along we judge that situation to be — not the probability it happens tomorrow. A high number means the conditions we said to watch for have largely occurred.
These are judgements, not measurements. They move when the evidence moves, and we show you the date each one was last reviewed so you can see how current it is. A number we have not looked at in months is worth less than a fresh one, and we would rather you know that than assume otherwise.
How strongly we hold a view on a specific company, from 0 to 100. It combines how far along the underlying situation is with how directly that company is exposed to it.
High conviction is not a promise. Our own track record — published on the Atlas Predictions page — shows roughly a third of closed trades were profitable, with the return coming from a few large winners. We would rather show you that than a flattering number.
Because it keeps us honest. If we say a chokepoint disruption is likely to last a month, that claim should come from somewhere you can check, not from a hunch we phrased confidently.
You never need to read the papers. They sit under a “where this comes from” link so the reasoning is auditable if you want it, and out of the way if you do not.
Every call is logged with the date it should be judged on and the condition that would make it right, before the outcome is known. On that date it is scored against the real price. Nothing is re-graded later and nothing is quietly dropped for going badly.
The scorecard uses a Brier score, which measures whether our confidence matches reality: it compares how sure we said we were against what actually happened. Below 0.25 beats saying “50/50” to everything. First results land in September 2026.
Public market data, delayed at least 15 minutes. That is a deliberate choice: real-time exchange feeds carry per-user licensing fees that cannot exist at $5 a month, and we would rather be upfront about the delay than pass on the cost.
No, and not in the fine-print sense — we are genuinely not set up to be. We do analysis and publish how it scored. We do not know your situation, your timeline, or your risk tolerance, and none of the numbers here account for them.
For educational and research purposes only. All analysis reflects the author's personal assessment grounded in published academic game theory. Not investment advice. Past frameworks do not guarantee future outcomes.