Industries / Finance and business services / Finance and banking / Energy trading
Physical trading
Buying, selling and shipping actual barrels, cargoes and gas volumes, and setting the prices at which they change hands.
6
Milestones achieved, from 1895 to 2026.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1895
Standard Oil sets the price
From 1871 the exchanges of Pennsylvania's oil region traded pipeline certificates, paper receipts for oil held in a pipeline, which traders bought and sold many times a day while the oil stayed put. On 23 January 1895 Standard Oil's purchasing agency in Oil City told producers it would pay a price "as high as the markets of the world will justify", not the price bid on the exchange. Its chief buyer, Joseph Seep, then paid producers directly at a price Standard set itself, and the exchanges lost their place.
Lessonwhen one buyer takes most of the oil, it can put its own price in place of the market's.
Credit
Standard Oil Company; the historian Brian Black and the American Oil & Gas Historical Society, whose accounts are used here
1986
Exporters price off benchmarks
After the 1979 oil crisis, producers stopped selling mainly through the large oil companies, and far more oil moved in short-term and spot deals. In 1986 Mexico's state oil company PEMEX began to price its crude by formula: the published price of a benchmark crude, plus or minus a set difference, in place of a fixed official price. By 1988, in the economist Bassam Fattouh's account, this had become the main method of pricing crude oil in international trade, and it still is.
Credit
PEMEX; Bassam Fattouh, Oxford Institute for Energy Studies
1988
Brent gets a futures market
The International Petroleum Exchange in London launched Brent crude futures in June 1988, "after a number of failed attempts". The contract settles against forward Brent, an informal market in North Sea cargoes for future months. By Fattouh's 2011 count, prices from this "Brent complex" directly or indirectly priced about 70% of international oil trade.
Credit
International Petroleum Exchange, now part of ICE (Intercontinental Exchange)
2001
Prices set in a window
Price reporting agencies are firms that publish the daily prices that physical oil contracts refer to. In January 2001 Platts stopped averaging the day's deals by volume and switched to "market on close", in which only bids, offers and deals made in a short window late in the trading day count. Critics say the window can be thin: Fattouh cites a study by Argus, a rival agency, which found that window trades were a small share of the day's volume in US crude in 2007.
Credit
Platts
2020
Invented sales sink Hin Leong
In April 2020 Hin Leong Trading, a Singapore oil trader, revealed more than US$800 million of losses it had not disclosed, by Global Trade Review's report. Its founder, Lim Oon Kuin, was first charged on 14 August 2020 and faced 130 charges of cheating and forgery against financial institutions. In 2024 a court convicted him of cheating HSBC: his staff had presented two oil sales, to China Aviation Oil and Unipec, that never happened, backed by forged papers, and the bank paid out US$111,683,939 against them. He was sentenced to 17.5 years in prison in November 2024.
Lessona bank that lends against trade papers needs a way to check that the cargo and the sale are real.
Credit
Singapore Police Force; the courts of Singapore
2026
War shrinks the Dubai benchmark
Dubai is the price used to value Middle East crude, and Reuters reports that it prices 18 million barrels of oil a day. On 2 March 2026, soon after war shut the Strait of Hormuz, Platts stopped counting crude loaded inside the strait, which cut the benchmark from five grades to two: Murban shipped from Fujairah, and Oman. Platts said this reduced deliverable crude by about 40%. On 19 March Dubai's front month was assessed at $166.80 a barrel against $125.90 for Murban, by Bloomberg's report, and some traders told Reuters the benchmark was "effectively broken".
Lessona benchmark works only while enough oil can be delivered against it.
Credit
S&P Global Energy Platts; reporting by Reuters and Bloomberg
Watch for
Dubai on two grades
Platts says the narrowed benchmark still reflects "the tradable spot market value of physical crude oil in the Middle East". It has asked the market how the method should change if the strait stays shut or reopens, and participants' ideas range from halting the assessment to adding grades from outside the region.
Achieved whenPlatts publishes a settled method and outside reporting shows Dubai-priced cargoes and Dubai derivatives trading normally under it.
Claimed by
S&P Global Energy Platts, March and April 2026
Open
01
Planted fakes in cargo papers
A public test set of about 500 made-up document packs for oil cargo deals, each with a sale contract, invoice, bill of lading and inspection certificate. Some packs carry planted frauds of the kinds seen in real cases: a sale the buyer never agreed to, one cargo financed twice with two lenders, a bill of lading copied from another pack, and quantities that do not add up. Each pack has an answer sheet, and a scoring script reports how many fakes a checking tool catches and how many honest packs it flags by mistake. Every party, ship and amount is invented.
The briefCredit · unclaimed02
The first oil prices, digitized
An open table of the daily crude oil prices printed in The Derrick's Hand-Book of Petroleum (Oil City, 1884) for 1875 to 1884, with the page for every row, so that the prices of the first oil exchanges can be studied like any modern price series.
The briefCredit · unclaimed
Open 01 · the brief
Planted fakes in cargo papers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test set of about 500 made-up document packs for oil cargo deals, each with a sale contract, invoice, bill of lading and inspection certificate. Some packs carry planted frauds of the kinds seen in real cases: a sale the buyer never agreed to, one cargo financed twice with two lenders, a bill of lading copied from another pack, and quantities that do not add up. Each pack has an answer sheet, and a scoring script reports how many fakes a checking tool catches and how many honest packs it flags by mistake. Every party, ship and amount is invented.
Why it is still open
Trade finance is "particularly susceptible to duplicate financing", the International Trade and Forfaiting Association says, and Deutsche Bank warns that routine checks of vessel movements "can deliver a positive response – even when the bill of lading presented to the bank is, in fact, incorrect". Small open-source checking tools appeared on GitHub in 2026, but my searches of GitHub and Hugging Face found no labeled test set of commodity document packs on which to score them.
Built with what exists
Python to generate packs from a list of fields; a PDF library and scan-noise filters for realistic copies; AI models such as Claude, GPT and Gemini, and open models such as Qwen, as the first tools to score.
First step
Write the field list for a crude oil cargo pack, then make 10 honest packs and 10 with one planted fraud each by hand.
Done looks like
500 packs, at least 100 of them with planted fakes across at least five fraud types, each type tied to a published case or bank warning. Catch rates and false-alarm rates are published for at least three tools, and a second person who re-labels 50 packs agrees on at least 49.
Open 02 · the brief
The first oil prices, digitized
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open table of the daily crude oil prices printed in The Derrick's Hand-Book of Petroleum (Oil City, 1884) for 1875 to 1884, with the page for every row, so that the prices of the first oil exchanges can be studied like any modern price series.
Why it is still open
The archive.org text of the book turns its price tables into garbled characters, and Weiner's 1992 working paper on the oil exchanges has no online copy. Searches of GitHub, Hugging Face and Crossref found no dataset of oil prices for this period.
Built with what exists
the book's 126 page images on archive.org, from a Library of Congress copy; an AI model that reads tables from images, such as Claude, GPT or Gemini; Python to check each row against the book's own table of highest and lowest prices at Oil City.
First step
Test two models on 10 pages of the 1875 market quotations against a hand transcription of the same pages.
Done looks like
every quoted trading day from 1875 to 1884 in the table, no row outside the book's monthly high and low, and a hand check of a random 5% of rows that matches the page images on at least 99% of numbers.