Methodology

Chokepoint Pressure answers “are the oil chokepoints in trouble right now?” Its base is the Strait of Hormuz, blended from two independent kinds of on-the-ground evidence, weighted by how directly each speaks to a disruption:

• Physical flow (60%), how many ships (tankers and other vessels) are actually moving through the strait versus a normal week, from satellite-tracked transit data.
• Conflict & jamming (40%), GPS interference in the region, a measurable side-effect of military activity near the shipping lanes.
• Bab el-Mandeb (additive, up to +25): since July 20, 2026, disruption at the Red Sea gate ADDS points on top of the Hormuz blend (scaled by how far its transits fall below their own recent normal). It is a sum, not an average: a healthy second gate contributes zero and can never water the gauge down.
• Oil market (0%, deliberately), Brent and the Brent–U.S. crude gap are still computed, but since July 2026 they no longer move this gauge: a calm oil market was muting the physical reality. They live on the market side of the divergence instead.

Early Warning answers “is market stress actually arriving?” It shows the average of the two hottestof six fast stress signals: how quickly credit spreads are widening, the Federal Reserve’s excess bond premium, options turbulence versus the turbulence markets actually delivered, what companies pay for short-term cash over the government rate, oil-option fear (OVX), and whether bond markets are demanding emergency rate cuts. Top-two rather than the average of all six, because one loud siren should never be muted by five quiet ones. Before it shipped we replayed it against every S&P 500 drawdown of 15% or more since 2006: it never once fired in a calm market, fired seven weeks ahead of the 2008 crisis, and days ahead (or none) elsewhere. Read it as a siren, not a forecast: it tells you stress is arriving, never that it will.

Market Fragility answers “if a shock lands, how hard does it hit?” It is the equal-weight average of six conditions that amplify any shock: the current inflation rate, the bond market’s inflation expectations, wholesale diesel, pump prices, how restrictive the Federal Reserve’s rate is in real terms, and how fast borrowed money in stock accounts (margin debt) is growing. In the same twenty-year replay it sat high through 2021 and 2022, ahead of that bear market, and near zero before the 2020 crash, which arrived in a robust market. It measures exposure, never timing: fragile markets can rise for months.

Two further scores run underneath without headlining the page. Market Transmission, “has the shock reached markets and your wallet?”, blends five indicators at equal weight (20% each): the U.S. Office of Financial Research’s daily financial-stress index, the VIX “fear gauge”, weekly initial jobless claims, pass-through to gas and diesel prices, and the Cleveland Fed’s daily inflation nowcast. It remains the market side of the divergence below and the embeddable badge’s fallback. The Recession signals score (the yield curve, the credit-spread level, and stocks below their 52-week high, with diesel and inflation expectations) watches the months-to-a-year horizon; its cards are grouped under that name on the dashboard.

We weight by independent kind of evidence, not by number of indicators. Transits, freight, and oil price all encode the same oil story, averaging them together would count it three times. Within each group, indicators are weighted equally.

Most indicators are anchored to two real reference points: a pre-crisis normal (scores 0) and a historical crisis extreme (scores 100). A reading in between is placed proportionally. Brent crude, for example, anchors at roughly $70 (its recent multi-year normal) and $150 (near its all-time 2008 high). This is deliberately not a rolling percentile: a percentile quietly redefines “normal” as a crisis drags on, so a months-long disruption starts scoring as calm. Fixed anchors don’t drift.

A few fast-moving market signals are the exception, because their “normal” genuinely shifts: the Brent–WTI gap and weekly jobless claims are scored as distance from their own recent baseline (the gap versus its trailing median, claims versus their lowest point over the past year). For those, the baseline is built to adjust quickly after a regime settles but slowly during a live spike, so a real crisis still can’t fade itself to calm by simply lasting a long time. Ship traffic’s “normal day” is likewise computed from its own history, the median of healthy days only, with a fixed floor, so neither a long disruption nor a short pre-closure rush of ships can redefine what normal means.

The 0–100 scale maps to four bands: Calm (0–25), Elevated (25–50), High (50–75), and Severe (75–100). In a genuine tail event, indicators can read above 100, we show that as an off-the-chart state rather than flattening the gauge at its maximum.

A few indicators aren’t a straight line and would be misread by anchoring, the yield curve (the pattern of interest rates across short- and long-term government loans) is the classic example, where the warning sign is short-term rates rising above long-term ones and then reversing, not any single value. Those cards are tagged context, not scored and never fold into a gauge.

An actual closure of the strait is the fact that matters, so the headline can’t be averaged down by a quiet news week elsewhere. When physical flow collapses into the “high” zone, Chokepoint Pressure is floored to high and scales toward severe as it approaches a full closure, even if oil prices and conflict signals are momentarily calm. If those other signals are pushing the score higher, the normal blend still wins.

This is the idea the whole page is built around. We score reality (what’s happening on the ground, ships, conflict, oil’s own signals) and the market (what financial prices have actually reacted to) on the same 0–100 scale, then watch the gap between them.

When they agree, there’s nothing to see. The case worth noticing is when the ground looks stressed but markets look calm, a gap that says the two sides are telling different stories right now. We show you the gap; we don’t predict which side moves, or when. We only flag a divergence once it has persisted (not on a single noisy reading), and if either side’s data is too stale to judge, we say so instead of implying agreement.

Separately from the gauges, we show the U.S. crude oil “days of supply”, the national stockpile divided by how fast refineries use it, published weekly by the U.S. Energy Information Administration. It’s a measure of slack: how much cushion exists before a real disruption would start showing up in prices. The dashed projection simply extends the recent trend a few weeks forward as a reasonable estimate, it is not a forecast.

Feeds lag, and some publish slowly by nature. Each indicator has its own staleness window, by default about twice its expected update interval, but wider for feeds with known publish delays (the ship-transit backbone, for instance, is kept for up to about two weeks). Past that window we drop the reading rather than pretend it’s current, and the gauge re-balances across the indicators that are still fresh. If too much of a gauge goes dark, under half its weight, we mark it “low-confidence” instead of showing a falsely precise band. Every card carries its own “as of” time so you can see exactly how fresh each piece is. This honesty is a feature, not a footnote.

There is no free data source that reliably tracks individual ships inside the Persian Gulf, the free public feeds of ship-transponder signals (called AIS, the system vessels use to broadcast their position) have no coverage there, so we don’t render fake ship positions. The vessels on the map are an illustration scaled to the real transit count, clearly labeled as such; the jamming heatmap, by contrast, is real per-location GPS-interference data.

We also leave out anything that would make this a forecast or a trading signal: no price targets, no “buy/sell,” no crash predictions. Jamming is used as a published proxy for conflict intensity, not a precise measure. Where a signal cuts both ways or can’t be cleanly measured for free, we show it as context or not at all.