The miss rate at Hormuz is six times higher than at Bab-el-Mandeb. Maritime risk teams underwriting Persian Gulf cargo should treat that as a coverage problem, not a routing one.

The Setup

Over the last 30 days, the chokepoint-transit pipeline emitted 1,294 predictions across the four monitored corridors. The Strait of Hormuz alone accounted for 1,196 of them.

Of those Hormuz predictions, 687 — 57.4% — were marked missed. A missed transit is a prediction where the vessel was inferred (via straight-line dead reckoning from its last AIS course and speed) to enter the chokepoint within an ETA window, and 96 hours later did not reappear in the exit-zone polygon.

The same pipeline tracking the same vessels at Bab-el-Mandeb produced a 9.1% miss rate. Suez South: 16.3%.

| Chokepoint | Predicted (30d) | Missed | Confirmed | % Missed | |---|---|---|---|---| | Strait of Hormuz / N Gulf | 1,196 | 687 | 0 | 57.4% | | Suez S approaches | 43 | 7 | 0 | 16.3% | | Bab-el-Mandeb | 55 | 5 | 0 | 9.1% |

The Chain

The prediction is mechanical: every hour, populate_predicted_transits() reads the most recent AIS position for vessels with course pointed at one of the four chokepoint polygons, dead-reckons an ETA, and writes a row to vessel_transits with status predicted. The reconcile-vessel-transits cron runs at minutes 20 and 50 of every hour. For each predicted row past its ETA + 96 hours, it asks one question: did the vessel's IMO show up in the exit-zone polygon? If yes, status flips to confirmed. If no, status flips to missed.

The 57.4% Hormuz figure is what happens after that 96-hour window has closed. It is not a measurement of vessels going dark — it is a measurement of vessels predicted to thread the strait southbound, ETA elapsed, and 96 hours later they are not on the south side.

The 9.1% Bab-el-Mandeb baseline tells you what model error looks like in a corridor where Persian-Gulf-internal port calls aren't a confound. About one in eleven predictions misses for routine reasons: course changes, speed shifts, vessels stopping early. That's the floor.

The Implication

The Hormuz miss rate is not a single signal. It is at least three superimposed signals:

Internal Persian Gulf port calls. A vessel predicted to transit Hormuz southbound from an inbound-side bearing may be heading instead to UAE, KSA, or Iranian ports inside the Gulf. From the model's vantage point, those are misses — they never appear at the exit.
AIS coverage thinning inside the Gulf. Several stretches of the upper Persian Gulf have spotty terrestrial AIS reception. A vessel that goes silent for 18–36 hours and reappears on the north side, then proceeds to a Gulf-internal port, contributes to the miss count.
Sanctions-evasion behavior. Some fraction of the missed transits are vessels deliberately turning off AIS while inside the Gulf. Without SAR corroboration (currently offline pending Copernicus credentials), this signal cannot be isolated from the other two.

For a maritime underwriter, the directional claim is: at Hormuz, AIS-only coverage is reliably wrong about destination at the high end of a single-decimal-percent range. Bab-el-Mandeb is a much cleaner channel. If your risk model treats both with the same prior, you are over-confident on Persian Gulf cargo by an order of magnitude.

What to Watch

Hormuz miss rate week-over-week. Rising rate signals either coverage degradation or evasion uptick. Falling rate signals a return to the long-run baseline.
Bab-el-Mandeb miss rate climbing above 15%. That would be the first quantitative indicator that Houthi-area effects are bleeding into AIS reliability.
The first non-zero confirmed count. The promotion path predicted → confirmed exists in the schema but is currently empty across all four chokepoints — the reconcile cron is firing but no exit-zone reappearance has registered.

Limitations

This is a 30-day window. The reconcile path needs at minimum ETA + 96h to mark a transit either confirmed or missed; the most recent predictions in the window are still pending. The 0 confirmed count across all chokepoints is the system's current upper bound on its own corroboration capability — until the SAR pipeline (Sentinel-1 CFAR detection via Copernicus) returns to active service, the only positive-confirmation signal is exit-zone AIS reappearance, and the polygons may be tighter than the actual transit corridor. The 57.4% Hormuz figure should be read as a coverage-and-routing measurement, not a sanctions-evasion claim. The signal converging with sanctions data requires the SAR confirmation path to come back online.

Data current as of 2026-05-07. Sources: Axiom Overwatch vessel_transits table, chokepoints polygons (Bab-el-Mandeb, Hormuz, Suez S, Cape Agulhas), predict_chokepoint_eta() function. Free public data from AISStream.io and Sentinel-1 (when available). Read more at https://www.axiomoverwatch.io/blog.