Commodity trading enters the age of digitisation
July 10, 2018 | 0 Comments
Commodity houses are on the hunt for data experts to help them gain an edge after seeing their margins squeezed by rivals.
Currencies, equities and interest-rates investors have for years used algorithms, machine learning and artificial intelligence to turn data into successful trades.
Now, commodity traders are seeking ways of exploiting their information to help them profit from price swings. “It is really a combination of knowing what to look for and using the right mathematical tools for it,” said Peter Leoni, who holds a mathematics PhD and is one of the two data scientists within a newly created team of 10 at EDF Trading, the London-based trading arm of the French utility group.
“We want to be able to extract data and put it into algorithms,” added Philipp Büssenschütt, EDFT chief commercial officer. “We then plan to move on to machine learning in order to improve decision-making in trading and, as a result, our profitability.” The French trading arm is investing in people, processes and systems to centralise its data — and it is not alone.
“Everybody [in the commodity world] is waking up to the fact that the age of digitisation is upon us,” said Damian Stewart at headhunters Human Capital. In an industry where traders with proprietary knowledge, from outages at west African oilfields to crop conditions in Russia, vied to gain an upper hand over rivals, the democratisation of information over the past two decades has been a challenge.
Through the broad dissemination of news, weather reports and cargo-tracking, commodity traders have found their margins under pressure as their information edge, that once buttressed their profits as middlemen, has been blunted. Return on equity for leading trading houses, a measure of the profit generated from the money shareholders have invested, has dropped significantly.
Some of the oil and metals traders enjoyed returns of about of 50-60 per cent in the mid-2000s, but this has declined to levels in the mid-teens. Agricultural traders’ returns have historically been lower, but they have also dropped, with the commodity companies known as the ABCDs — Archer Daniels Midland, Bunge, Cargill and Louis Dreyfus Company — all recording single-digit ROE in their latest results.
As a consequence, an increasing number of traders are hoping to increase their competitiveness by feeding computer programs with mountains of information they have accumulated from years of trading physical raw materials to try and detect patterns that could form the basis for trading ideas.
“In agriculture, metals or energy, the traders are looking to gather data on a large scale and run machine-learning algorithms to find patterns linking fundamentals with price movements,” said Etienne Amic, a former JPMorgan banker and chairman of Vortexa, a company making cargo-tracking software.
Cargill, for example, started to build its global digital team two years ago and now has 75 people focused on digital innovation and incubation, with a data science team of 12. According to Gert-Jan van den Akker, president of Cargill’s agricultural supply chain division, the combination of data and cutting-edge technology is already leading to better trading decisions.
“Humans have always played a vital role in trading and understanding futures markets, but we’re no longer relying on human brain power alone,” he said. At the same time, computer models and algorithmic programs are also exerting greater influence on the prices of commodities in the futures markets, making it more difficult to hedge positions.
“We notice that the fundamentals and prices increasingly get out of balance in the market with the increased volumes executed by algos,” said Mr Büssenschütt of EDFT. Between 2012 and 2016, almost two-thirds of crude oil contracts traded on CME’s futures exchange were automated, up from 54 per cent, according to a 2017 study by the US Commodity Futures Trading Commission. In soyabeans and wheat, the figure rose from 39 per cent to almost half, while in precious metals it has climbed to 54 per cent from 46 per cent.
Some traders are joining forces with algorithmic funds to learn how they work. BayWa, a German midsized agricultural trader, has linked up with Molinero Capital Management, a US fund specialising in computer-backed “quantitative” trading, and German fund Quantumrock Capital. Despite this new enthusiasm, the road to electronification may not come easily for some traders. Compared to other financial and industrial sectors, “they are coming from way behind,” said one consultant. One issue is that some of the larger commodities traders face internal resistance in centralising information on one platform.
With each desk in a trading house in charge of its profit-and-loss account, data are closely guarded even from colleagues, said Antti Belt, head of digital commodity trading at Boston Consulting Group. “The move to ‘share all our data with each other’ is a very, very big cultural shift,” he added. Another problem is that in some trading houses, staff operate on multiple technology platforms, with different units using separate systems. Rather than focusing on analytics, some data scientists and engineers are having to focus on harmonising the platforms before bringing on the data from different parts of the company.
Even where the digital infrastructure is in place, it may take some time before AI becomes a large part of commodities trading. Vitol’s chief information officer Gerard Delsad has increased his team by a fifth to more than 100 in the past three years, including several data scientists. He said a computer still struggles to find patterns in the data and come up with trading ideas on its own. “You still the need the trader to give some ideas, some hints [of] where to look, and then you can find some interesting stuff.” Nevertheless, digitisation is increasingly driving trading and needs to be embraced, say many commodities executives. Mr Büssenschütt said: “It’s another tool that traders have to understand.”