About · forecasting in public, every call graded

One human, one number, one scorecard.
Calibrated forecasting on where geopolitics moves markets — and it grades itself.

Every call gets a number, a resolution date, and a public score when it resolves. No vibes, no “tensions are rising” — a probability, a deadline, and a scorecard.

I’m Chris. I built the engine, I make the calls, I eat the misses. Every call I make is logged, timestamped, and graded in public at /receipts — you never have to take my word for my accuracy. You can check it.

The author · I own the calls and the misses
Chris

The forecasts are amplified by an intelligence engine I’ve spent months building — 70+ news feeds, Telegram OSINT, financial market tells, seismic data, and superforecaster consensus — so the weekly read is built on a far wider base than one person could watch alone. But the call, and the accountability for it, is mine. When I’m wrong it goes at the top of the next issue, not down the memory hole.

What this is

A number, a deadline, and a scorecard.

Most geopolitics writing tells you a confident story and never checks whether it came true. Most market commentary ignores the geopolitics driving it. I do both, with numbers. Each week I put explicit probabilities on resolvable questions — Will the US and Iran sign a nuclear deal this year? Does a Hormuz scare push Brent above $80? What does the Fed do in June? — say where I disagree with the market and why, and then let you watch me be right or wrong in public.

The edge is the seam between the two worlds: I read the conflict signals and the instruments that react to them — oil, gold, defense, the dollar — so the call isn’t “tensions are rising,” it’s “here’s the probability, and here’s what it means for the tape.”

Why it's different

Three things almost no geopolitics newsletter does.

There are thousands of geopolitics newsletters. Almost none of them do these.

1. Put a falsifiable number on anything.

“Escalation is likely” is unfalsifiable. “31% by August 1” is a bet I can lose.

2. Keep score in public.

Every call I make is logged, timestamped, and graded when it resolves — track record live at /receipts. You never have to take my word for my accuracy. You can check it.

3. Own the misses out loud.

When I’m wrong, it goes at the top of the next issue, not down the memory hole.

How the scoring works

Calibration is a real, measurable skill.

Every call is a probability (0–100%) on a question with a hard resolution rule and date. When it resolves, it’s scored with a Brier score — the standard measure of forecast accuracy (lower is better; a coin flip is 0.25). The running track record — every open and closed call — is public at /receipts.

The point of this newsletter is to prove that skill on the only thing that matters: the record. And I won’t overclaim it before the data can carry the claim. Only ~4 distinct markets have resolved so far — the skill number is the open record, not a finding; we’d need ~200+ resolved questions to tell skill from luck. So the honest read today is “underpowered — can’t demonstrate edge yet,” and the fast-clock markets calls are how the record gets there. You can watch the count climb on the receipts page.

The engine behind the call

One person, watching far more than one person could.

The engine is the muscle. The call, and the accountability for it, is mine.

Every cycle, the engine reads twelve intel layers — wire news, native-language press, OSINT Telegram, Twitter, financial instruments (Brent, gold, the VIX, the dollar, defense), prediction-market order flow, smart-money positioning, USGS seismic, the scheduled macro calendar — across 119 news feeds in several languages. It tags each item to a typed entity graph and scores every tracked market, so the weekly read sits on a far wider base of evidence than any one person could watch alone.

But none of that is the author. The engine surfaces and grounds; I decide the number, write the thesis, and own the result when it resolves. If you want the full mechanism — how items get grounded, how each market is scored, how the record is graded — it’s all on the methodology page.

Cadence & price

Weekly, plus a flash when something breaks.

Weekly, plus a short flash when something actually breaks.

Free while the track record builds. If it earns it, a paid tier comes later — and you’ll have a public scorecard to decide whether it’s worth paying for. That’s the whole idea.

Who writes it

One human, owning the record.

I’m Chris. I built the engine, I make the calls, I eat the misses. One human, one number, one scorecard. In a category full of confident, unfalsifiable takes, the differentiation is rigor and a person who owns his misses in public — and that’s the part no engine can copy.

Calls are forecasts, not advice. Every call is a probability on a resolvable question, graded in public when it resolves. The track record is still young — only ~4 distinct markets have resolved, far short of the ~200+ needed to separate skill from luck, so treat today’s record as the open ledger, not a proven edge.
Not financial advice. Prediction-market trading involves significant risk of loss. Past performance does not guarantee future results.