Algorithmic trading in energy markets

Algorithmic traders are now a significant presence in energy markets. Alexander Osipovich explores how they’re changing the game for human traders

They made their debut on the stock market, and gained notoriety after the ‘Flash Crash’ of May 6, 2010. Today, algorithmic traders who use computers to execute thousands of trades in seconds have become major players in energy markets, too.

And that has sparked debate among long-time market participants over whether the high-tech newcomers are to be welcomed or feared.
Algorithmic trading comes in many forms, but it can broadly be defined as the use of programs to act on market data and make trades without human intervention.

High-frequency trading (HFT) is a subset of algorithmic trading in which trades are executed at lightning speed, usually to take advantage of fleeting arbitrages between different markets. Critics of algorithmic trading believe it’s altering historical correlations between markets, making commodities move in tandem with equities and negating their use as a portfolio diversifier. They also point to incidences of rogue algos that can undermine normal market behaviour and even cause flash crashes.

Algo traders and HFT firms can only operate in highly liquid markets, so when it comes to energy, they tend to favour heavily traded contracts such as WTI and Brent crude oil futures or Henry Hub natural gas futures.

“We trade WTI, and we trade gas futures on Nymex,” says Karel Janeček, the chief executive and biggest shareholder of RSJ, an algorithmic trading firm based in Prague. Energy is a modest part of RSJ’s business, though, and the lion’s share of its revenues comes from market-making in financial futures, says Janeček. “Liquidity is the key consideration. The more liquidity, the more interesting the market for us. Also, the other thing is the cost issues. The fees in energy are relatively high compared to short-term interest rate futures.”

About 45% of trading on Nymex is algorithmically driven, of which a smaller fraction is true high-frequency trading, CME Group chief executive Craig Donohue was quoted as saying by Reuters earlier this year. By comparison, analysts have estimated as much as two-thirds of volume in equities is driven by algorithms and HFT. A CME Group spokesman declined to give more recent figures on the levels of algo trading, citing “competitive reasons”. Ice, the other major exchange for energy futures, has never released such data.

Many algo traders operating in energy markets use the same strategies that have long been used by commodities traders, but at much greater speeds and in much tighter time frames, according to analysts and market participants. For instance, they may develop a model of calendar spreads in crude oil futures and implement a mean-reversion strategy to make profits when reality diverges from their model’s predictions. “My understanding is a good deal of high-frequency trading in the commodities space is spread trading,” says Howard Tai, an analyst with Aite Group.

“Before the arrival of electronic trading, most energy trading was done by individuals in trading pits or bilaterally over the phone. The advent of electronic trading in energy allowed HFT firms to take part in energy markets. These high-frequency traders can do those spread trades much faster.”

What effect has this had on energy markets? Seasoned market observers say moves in the futures curve, which previously took some time to play out, now happen almost instantly. More broadly, the growth of algorithmic trading has increased the correlation between energy prices and key financial markets such as equities and the US dollar, argues oil analyst Olivier Jakob of Petromatrix.

“With the move to electronic markets and being able to run models, what we’ve seen is much greater correlation between markets,” Jakob says. That has frustrated investors who piled into commodities in recent years, hoping they would be uncorrelated with stocks. “Many pension funds I talk to today complain that there’s little diversification, because everything moves one way,” Jakob says. “You can be invested in commodities, but it doesn’t bring any diversification benefit.”

One aspect of algorithmic trading that has drawn criticism is the emergence of ‘predatory algos’. Such algorithms will typically attempt to sniff out a major transaction being undertaken by another market participant, such as a financial institution unloading a big position in a stock, then seek to profit from that knowledge – a strategy some have compared to front-running, which is illegal.

Predatory algos first appeared in the equities world, but they now prowl energy markets too, according to traders. “If you trade, you can actually see them working,” says Drew Wozniak, vice-president of market research and development at Icap Energy. “The market will be in a lull, it’s not really moving around a lot, then if you place a trade, suddenly things start moving around. So whatever predatory behaviour the algo has, you can actually see them at work.”

This has led to something of an arms race, in which market participants who do not want to be preyed upon by predatory algos have developed their own algorithms to make big transactions more stealthily. For big players in energy, another option is just to do things the old-fashioned way, and to call up a broker who can execute the transaction without having it pop up on the predators’ radar screens.

Algos gone wild

The arrival of algorithmic trading has led to occasional moments of extreme market behaviour, including spikes or sudden nosedives that seem to occur without rhyme or reason. One such incident took place on June 8, 2011. That day, at 7:39pm in New York, a time when the market is usually quiet, the price of Nymex natural gas futures for July delivery began to oscillate up and down in a pattern resembling a sine wave. The oscillations grew steadily greater, until 7:42pm, when the price abruptly plunged more than 37 cents – or about 8% of its value — in less than two seconds (see figure 1). The market recovered from this ‘mini-Flash Crash’ over the next 10 minutes or so.

F1. Oscillating Nymex gas prices shook the market on June 8 (Source: Nanex)

“That was really bizarre,” recalls Eric Scott Hunsader, founder of Nanex, a software company based outside Chicago that records market data and maintains a blog about strange phenomena purportedly caused by algorithmic trading. What was especially perplexing about the June 8 anomaly, Hunsader says, is that the waves were accompanied by counterintuitive behaviour in the depth-of-book data. “I’ve never seen something that crazy, because it just doesn’t make any sense at all. As the market was going up, the depth of book on the bid side was actually dropping, then the offer side was opposite, it was increasing. It’s almost like somebody put fake orders in there while it was oscillating up and down to fool other algorithms.”

Hunsader, a critic of high-frequency trading, has also highlighted behaviour in which he believes algo traders are making simultaneous moves in markets for popular commodity exchange-traded funds (ETFs) and for the underlying futures contracts, such as the United States Oil Fund (USO) and WTI crude oil futures. Nanex has observed such co-ordinated actions taking place just seconds before the release of a major report, such as the US Energy Information Administration’s weekly report on oil inventories.

“We’ll see a massive move in the ETF at literally the same millisecond as in the futures,” Hunsader says. What are these algos trying to accomplish? Hunsader says he can only speculate, but believes they are seeking to take advantage of other arbitrageurs with less speedy technology.

There has been at least one other documented case of an algorithm going rogue in the energy markets. In November, CME Group imposed a $350,000 fine on Infinium Capital Management, a Chicago-based algorithmic trading firm, for an out-of-control algo that had triggered a $1 spike in crude oil futures on February 3, 2010.

Infinium neither admitted nor denied wrongdoing, but said it had greatly improved its testing and risk management procedures since the incident. “We remain committed to facilitating orderly markets and establishing minimum industry standards so that similar errors are not brought to the marketplace by ourselves or other trading firms,” Infinium says.

RSJ’s Janeček says such incidents are rare and that algo traders themselves suffer the most when their algorithms go bad. “Who is hurt?” he asks. “If there is a spike in price and a faulty algorithm makes a bad sell or a bad buy, then it’s the owner of the algorithm who loses money. But it doesn’t have any impact on long-term position traders.”

HFT under fire

The events of May 6, 2010, in which the Dow Jones Industrial Average plummeted around 1,000 points, only to regain much of the lost value within a matter of minutes, sparked concern that algorithmic and high-frequency traders were adding new volatility to financial markets. But a joint investigation into the Flash Crash by the US Securities and Exchange Commission (SEC) and Commodity Futures Trading Commission (CTFC) largely cleared algo traders and HFT firms of responsibility. The joint SEC-CFTC probe found a mutual fund firm’s mishandling of a giant sell order, which dumped $4.1 billion worth of stock index futures contracts on the market in just 20 minutes, during a day that had already shown unusual volatility, triggered the freefall.

The probe also found, however, that high-frequency traders contributed to the slide by withdrawing from the market when things got bad, causing liquidity to dry up swiftly. This has led to criticism that HFT provides liquidity – except when it is most needed. “In general, it’s good. I think it creates greater efficiency,” says Michael Cosgrove, managing director for strategic initiatives in commodities and energy at GFI Group. “The problem, of course, is that they’re not position takers, and so as quickly as they get into something, they would like to get out of it.”

Algo traders and HFT firms deny they are destabilising markets, and point to several academic studies that have found no link between volatility and algo trading. “If anything, algorithmic trading has reduced volatility. It definitely has not increased it,” Janeček says. “Algorithmic trading has been very beneficial for the markets. It provides a lot of liquidity. We have much more liquidity, we have lower spreads, and that’s why hedging and trading for customers is much cheaper.”

Moreover, CME Group and Ice say they have put in place risk management controls to prevent runaway algos from causing abrupt spikes or drop-offs in prices – which would theoretically prevent Flash Crash-style scenarios from playing out in oil or gas markets.

Still, more traditional market participants remain wary. For instance, power-generation companies trying to assess risk may be thrown off when algorithmic trading causes unexpected moves in fuel markets, says Art Altman, a researcher in energy risk management. “At a minimum, it’s an additional layer of complexity that an analyst at a power company has to consider – one that may cause market behaviours to deviate from fundamentals as well as historical norms,” Altman says.

“For example, the risk manager may base hedging decisions on traditional seasonal trends in gas price volatility and/or correlations between gas and peak power prices, based on their understanding of underlying supply-demand patterns, as well as observing market behaviour over time. High-volume algorithmic trading has the potential to alter such patterns seemingly overnight.”

In the wake of the Flash Crash, the CFTC has been discussing the possible imposition of new rules aimed at supervising algo traders and HFT firms. But that process is still in its infancy, with commissioners and their advisers still discussing how to define the term ‘high-frequency trading’. In the meantime, energy market participants will simply need to get accustomed to the presence of algo trading. “It’s just created a new playing field,” says Wozniak, of Icap Energy. “And that’s okay. Markets evolve.”