A proprietary estimate of how much crude oil, condensate and petroleum products are moving through the Strait of Hormuz on a given day, expressed in million barrels per day, built only from inputs we are legally permitted to republish.
Two cleared datasets, both from one publisher: the U.S. Energy Information Administration (EIA). Both are works of the U.S. federal government and in the public domain. Relying on one publisher is a real limitation — if EIA’s figures share a blind spot, so do ours — and we say so rather than implying a wider net.
The anchor is EIA’s Global Energy Security Data, released 12 August 2026, covering quarters through April–June 2026. It is the only cleared series that measures the strait itself. We read two of its tables. Table 4 is the Strait of Hormuz. Table 2 carries the world’s other maritime chokepoints, and we use it as a control rather than as an input: the Danish Straits, the Turkish Straits and the Panama Canal carry no Gulf barrels, so when Hormuz moves sharply and those three do not, we can tell the change is specific to this strait and not a worldwide slowdown. EIA’s volumes are based on Vortexa tanker tracking data with additional EIA analysis; we republish EIA’s published analysis and do not hold or license Vortexa’s underlying data.
The monthly signal is EIA’s Short-Term Energy Outlook, Table 3d (September 2026 edition), which reports crude oil production by country every month, currently observed through August 2026 — two months past the end of the anchor. We add up five Gulf producers that depend on the strait (Saudi Arabia, Iraq, Kuwait, Iran and Bahrain) and use only months EIA has observed, never its forecasts. This measures production, not the strait: oil can go into storage, into local refineries, or into pipelines that bypass Hormuz altogether, which is exactly why it moves our estimate but does not replace the anchor. The United Arab Emirates and Qatar are left out because the same table reports them on a different basis, and mixing the two would quietly distort the total.
EIA notes that since the end of February 2026, AIS signal data for ships transiting the strait has become especially unreliable, and that 2026 Hormuz volumes are being revised frequently as better information arrives. We treat the anchor as provisional for exactly that reason.
Several daily trackers exist. We have checked their terms and rejected them: IMF PortWatch and the WTO/AXSMarine Hormuz tracker are both closed to redistribution, and so is the JODI monthly oil database, whose terms reserve all rights. We do not use an uncleared source “just to calibrate” — if it moves the number, it needs a licence.
Two numbers do the work. The first is the strait’s share: in the latest quarter EIA has measured, how much oil passed through Hormuz relative to what those five Gulf producers pumped. The second is the most recent month of their production. Our estimate applies the one to the other: it assumes the strait still carries the same share of Gulf output as in the last measured quarter, and lets the figure move month by month as that output rises or falls. That share is our main assumption between quarters, and it is the one most likely to be wrong — in a disruption it moves, and we cannot see it move until the next quarter is published.
The range is not a guess about a guess. We tested the method against the published record, predicting each quarter from the one before and using only the production months that would actually have been available at the time. While flows were steady it stayed within 2% of the measured figure. Through the 2026 collapse it overshot by 45–51% in the first quarter, no better than simply carrying the previous quarter forward (45%), and by 84–94% in the second, where carrying forward overshot by 204%. So in a disruption one end of our range is “nothing has changed since the last measured quarter”, and the other is the method’s largest measured miss, applied in the direction production has moved since then. That is why the headline does not sit in the middle of the range. When flows are steady the range closes to about ±3%. A mechanical test decides which regime we are in: anything beyond ±10% quarter-on-quarter is “disrupted”. Every time a new quarter publishes, the share resets, the estimate we published is scored against the measured figure, and that result joins the record the range is derived from.
That back-test only tells us how wrong this method gets one quarter past its anchor. It says nothing about how wrong it gets two or three quarters past, because we have never been there and measured it. So the model carries a hard limit: once the day being estimated is more than about one quarter beyond the end of the latest published quarter, we stop showing a daily number and say so plainly, instead of quietly stretching a range that was never tested that far.
Our current anchor covers April to June 2026, so that limit falls at the end of September 2026. If EIA has not published a newer quarter by then, this page will show no headline figure until it does. We would rather answer “not right now” than answer confidently and be unable to say why we believed it.
The method is the product, and it is not finished. Expanding the set of inputs — additional public datasets, official reports, credible reporting — is standing, continuous work. Each addition has to clear the same bar: we must be legally permitted to use and republish it. When the set changes, this page changes with it.
The full working method, including the back-test and every source checked and rejected on licensing grounds, is in the public project repository.