Throughout most of the 20th century, robot traders would have been a mere figment of the (sci-fi film-influenced) imagination. But now they’re used by over 80 percent of trade markets, including the majority of investment banks and other big institutions, with retail trading remaining one of the only sectors still reliant on human brains. In short, the conventional trader depicted in clichéd Hollywood movies has almost completely died out, and with it the elements of risk-taking and intuition that defined the industry for hundreds of years.
Attempts at automated trading aren’t new – the Black-Scholes formula, taken up by traders when it was revealed in 1973, is proof enough that traders have long been trying to predict share prices in the most accurate and profit-reaping way possible. But computerised traders have taken it to the next level.
With these robot traders you’d expect the element of corrupt trades and dodgy dealings to disappear, but a lack of transparency continues to plague the high-frequency trading (HFT) industry, and the speed at which computers operate makes tracking individual trades nigh-on impossible without an audit trail.
Rise of the machines
Eric Hunsader from US data firm Nanex believes robot traders fiddle the market, ordering then cancelling trades just before the critical buying moment. “If the regulator fully understood what the computerised trader was doing, it wouldn’t be legal,” he told World Finance. A recent case saw trading firm HTG Capital Partners accuse Allston Trading of that very activity, known as spoofing.
Nanex reported that the case marked the first time two large trading companies have been in a spoofing dispute, which led the CME, on which the exchange was made, to review its regulations. A source told Bloomberg that HTG believed the CME’s self-match prevention system (which stops a company carrying out a trade with itself) may have been exploited in spoof transactions, leading the CME to add that misuse of its software was a “violation” against trade regulations.
If the regulator fully understood what the computerised trader was doing, it wouldn’t be legal
In 2013, owner of Panther Energy Trading Michael Coscia found himself forking out $4.5m to regulators following an accusation concerning spoofing in commodities trading, and earlier this year co-owner of Visionary Trading Joseph Dondero and others involved were obliged to pay $2.5m to resolve claims they had carried out layering as well as spoofing. The potential for corruption in computerised trading is thus more than a mere myth, and it’s raised concern among various parties – not least Hunsader, who claims such trickery occurs every day. “Knight Capital and Citadel make fortunes from this obfuscation,” he says. Concern over the dangers of robot traders has led others to probe, including American Attorney General Eric Schneiderman, who is investigating their potential for manipulation.
Flash Crash
On 6 May 2010, the famed Flash Crash hit the stocks of the New York Stock Exchange, with a number of shares from key players, including Accenture and Procter and Gamble, plummeting to almost nothing, and others such as Apple surging to sky-high values of up to $100,000. The Dow Jones Industrial Average dropped by nine percent – 1,000 points – in the space of minutes.
A high-profile probe by the SEC found that computerised traders were behind the decline. “HFTs began to quickly buy and then resell contracts to each other – generating a ‘hot-potato’ volume effect,” the report read. “Between 2:45:13 and 2:45:27, HFTs traded over 27,000 contracts, which accounted for about 49 percent of the total trading volume, while buying only about 200 additional contracts net.”
According to Nanex, those moves might have been premeditated attempts at manipulation, although some, including the SEC report, refute the idea. Either way the circuit breakers put in place to prevent such shock incidents failed to act – a worrying indicator of their fallibility.
Hunsader says HFTs sometimes, somehow, work outside of the five to 10 percent parameters set by programmers. Since the so-called robots simply work by algorithms detecting differences in expected and actual stock prices, that shouldn’t happen. “There should be no decision making process, it should be very cut and dry,” says Hunsader. The Financial Conduct Authority (FCA) is only too aware of the frequently shady activity in computerised trades. “Unfortunately the nature of markets is that there always is potential for abusive activity, and with very, very fast trading, these things can happen very, very fast,” FCA boss Martin Wheatley told the BBC, adding that it’s then extremely difficult to detect the bad trades.
The Flash Crash isn’t the only example of robot traders gone wild; last year, in the space of 45 minutes, financial services company Knight Capital lost over $440m as a result of freak algorithm activity. Those blunders are likely to continue unless systems and regulators improve, according to Hunsader. “I can say we’re definitely going to get another sudden collapse in the market and it wouldn’t surprise me to see lots of stocks hit the circuit breakers,” he says.
Major firms are reluctant to implement that transparency, however, for fear other companies could copy their transaction patterns. US firm Virtu, one of the world’s biggest computerised trading software producers, indeed cited “extensive” scrutiny regarding regulation as one of the major disadvantages of its planned IPO earlier in the year. Head trader at US firm NorthCoast Asset Management, Frank Ingarra, agreed. “It could hurt them by opening them up to more scrutiny and regulation,” he told the BBC. “There’s just a lot of ambiguity and not a lot of regulation in that area of the market,” he added.
Hunsader believes computerised trading has potential to be “great” if regulation of HFTs does somehow come into force. That seems to be the view of Wall Street trader-turned-Cambridge University neuroscientist John Coates, who explores the risk-taking element of trading and its physiological effect in his book, The Hour Between the Dog and the Wolf. He writes that the biological response to risk-taking impairs human judgement, causing jumps and crashes in the stock market. Computers should theoretically be able to stabilise that, evading the problems human activity entails – but incidents like the Flash Crash suggest the contrary.
Insufficient replacements
What robot traders do evade are the human-specific elements that have for so long been fundamental – and beneficial – to trading. “In the old days, 10 years ago, a desk of equity traders would have between 80 and 100 of human traders at an investment bank. Today there’s maybe eight of them left,” Remco Lenterman, Director of technology trading company IMC Financial Markets, told the BBC. As human trader control wanes (and IT personnel monitoring the algorithms take over), so too does conscious risk-taking, decision-making and intuition, which computers simply cannot mimic. “They’re leeches on the system, they’re not contributing at all,” says Husander.
That lack of thinking capacity can add to the dangers; it was an absence of decision-making ability that saw the robots all suddenly withdraw from the Flash Crash (in reaction against the plunges), in turn sparking the stock spikes. And their inability to process and react to changes which might affect share prices in the way a human could means they can instigate significant losses. A sudden announcement that could transform a company in a matter of seconds and see share prices crash would go completely over the robot trader’s metaphorical head.
Eradicating humans also means reducing the diversity of traders. If one major HFT producer (such as Virtu) were to monopolise the market, then just one system would be responsible for all trading. That monopoly wouldn’t be unlikely given that the tech industry tends to be dominated by a few major players. According to Husander this too, would be detrimental to trading. “The most dynamic marketplaces, the ones most resilient to sudden shocks in the system are the ones that have wide diversity of participants,” he says.
Finding a solution to the problems entailed by mechanising trade isn’t easy. A return to human-only trading is hardly viable given the difference in profits robots can generate in comparison to human traders; the fastest human cognitive processing takes around 200 to 300 milliseconds according to Coates, while an HFT can process at around a millionth of a second. Achieving hundreds of million trades a day, robot traders’ minor profit margins soon add up, and they dwarf those achieved by humans – that’s provided their predictions for share prices aren’t proven wrong at the critical buying moment.
Husander believes the future of trading lies in combining human intuition with computer processing capacity, via an almost cyborg-like interface that would allow humans to input information at a rapid rate. Given the potential for fraudulent activity already apparent in robot and human traders, that’s a somewhat frightening prospect, and regulating it would pose an even greater challenge.
Regardless of where trading goes from here, robots have and will continue to transform its nature by replacing human thought. What’s apparent is that computerised trading needs to be brought out of the mystery currently shrouding it if progress is to be made. The problem is that making computerised trading more transparent – and unveiling the illicit activity it entails – is exactly what some traders don’t want.
