Methodology

The thesis

Presidential rhetoric and diplomatic headlines are noisy signals — they are cheap to produce and often decoupled from what the parties actually do. Hormuz Signal instead watches the physical and financial plumbing of the conflict: whether commercial shipping still moves through the Strait of Hormuz and the Red Sea, whether China keeps buying discounted Iranian crude, and whether global oil and refined-product inventories keep tightening. These are costly signals — shippers, underwriters, and refiners reveal real expectations through where they send hulls and how they price risk.

How the score is computed

1. Each indicator's latest value is converted to a raw stress reading relative to a peacetime baseline. For up-bad indicators (prices, rates, premiums, event counts) stress is (value − baseline) / span; for down-bad indicators (transits, imports, inventories) it is (baseline − value) / span.
2. Raw stress passes through a saturating curve n = 100·s/(s+0.5): 0 means calm, and 100 is an asymptote that is never reached — so even unprecedented stress keeps some headroom (see "Model v2" below).
3. The composite score is the weighted average of the normalized stresses. Indicators with no current data are excluded and the remaining weights are renormalized to sum to 1, so a missing feed never silently drags the score toward zero.
4. The result is rounded to one decimal and mapped to a risk level.

Model v2 (September 2026)

The original model mapped stress linearly onto 0–100 with a hard cap at 100. During the 2026 blockade several indicators hit that cap simultaneously, so the composite could no longer distinguish "severe disruption" from "unprecedented" — an active blockade read within a few points of open exchanges of fire. On 27 September 2026 the normalization was changed to a saturating curve, n = 100·s/(s+0.5), over unbounded stress, and the level thresholds were recalibrated (15 / 30 / 50 / 70).

• Nothing else changed: the indicator set, weights, baselines, spans, and staleness rules are identical.
• Under the new curve 100 is an asymptote — reserved, in effect, for genuinely unprecedented combinations. Everyday severe regimes sit lower, which restores separation between "bad" and "unprecedented".
• Scores before and after this date are not directly comparable: a given situation reads roughly 10 points lower on the v2 scale. The level bands (Low…Severe) were recalibrated so that the meaning of each band carries over.

Model v2.3 (October 2026)

The composite gained a force-posture cluster: US carrier strike groups and amphibious ready groups in Iran-relevant waters (weekly USNI Fleet & Marine Tracker as ground truth, with movement news flipping per-hull state between weekly posts), and military tanker/airlift tempo from open ADS-B. The model previously had no visibility into military pre-positioning — yet posture is escalation risk independent of intent: miscalculation and accident risk scale with forces in theater, and deployments are observable physical facts, not rhetoric. Existing weights were re-carved to fund the cluster (0.16 of the composite) and now sum to exactly 1.

Risk levels

Indicators

13 series, weighted by how directly they measure escalation pressure. Baseline is the peacetime reference value; a move of one span in the direction of risk maps to 66.7 stress. Larger moves approach 100 without reaching it.

Caveats

• Several series (tanker rates, war-risk premiums, Chinese imports of Iranian crude) have no free real-time public feed; they rely on proxies or periodic reporting and may lag events by days.
• Weights, baselines, and spans are analytical judgments, not fitted parameters. The score is a structured way to read the data, not a forecast.
• A 0–100 score implies false precision. Treat the level band and the trend of the components as the signal, not the decimal.