As machines take over the world of investing, data analysts and coders are fast becoming the new rulers. From Brainstorm magazine.
Michael Lewis is best known as the quirky fund manager depicted playing drums in his office in the movie The Big Short. He predicted – a little too early – that the US housing market was headed down the drain, and managed to make a fortune betting against the banks. It became clear to Lewis and others that the banks were recklessly doling out mortgage loans, often to people without jobs, and then bundling these loans together and off-loading this junk as AAA-grade mortgage-backed bonds to big institutional investors. It was one of the great scandals of the early 21st century.
Lewis’ next opus was a book called Flash Boys, which explores the relatively modern phenomenon of high-frequency trading (HFT). HFT, as defined by Investopedia, is ‘a method of trading that uses powerful computer programs to transact a large number of orders in fractions of a second. It uses complex algorithms to analyse multiple markets and execute orders based on market conditions’.
Lewis describes how a US company called Spread Networks invested $300 million building a 1 330km cable from Chicago to New York with the aim of reducing data transmission latencies from 17 to 13 milliseconds. That was in 2010, and marked a huge improvement in transmission speeds available at the time. In 2011, Hibernian Atlantic announced that it was building a transatlantic cable between London and New York to shave five milliseconds off transmission speeds.
There is a limit to how fast these speeds can get, and that limitation is the speed of light. The only way to increase speeds is to shorten the cable, which is exactly why HFT firms started taking up office space just metres away from the stock exchange computers. Being so close to the action means they can execute trades almost at the speed of light.
Those closest to the stock exchange engines are first to receive data, process it through their algorithms, and then execute trades. It’s reckoned that HFTs account for up to two-thirds of all stocks traded in the US, and perhaps a third on the JSE.
Closer to home
“It’s going to be much more difficult going forward for high-frequency traders to make the kind of money they made in the past, so intelligent algorithms are going to have to find new ways to identify profit-making opportunities.”
Fanie Harmse, FX & Project Management
In 2013, the JSE launched its colocation services, which allows traders to position themselves within shouting distance of the market infrastructure. It boasts a round-trip co-location latency of less than 100 microseconds (0.1 milliseconds) and, to all intents and purposes, bypasses the telco providers. To level the playing field, all cables to the market engine are of equal length, and time synchronisation ensures everyone receives the same information at the same time.
The JSE won’t divulge any stats around HFT – in fact, the algorithms behind HFT are as close to top-secret as you’ll find in the financial markets. Those in the business aren’t saying much, but we can get a sense of where this is going from the JSE’s colocation stats (not all of these trades would be HFTs).
The ‘flash crash’ of 2010, which was the second worst intra-day drop in stock exchange history – a trillion-dollar crash affecting multiple stock markets and lasting 36 minutes – was partly blamed on high-frequency traders. That said, regulators agree other factors were also at play, such as a mistakenly large sell order for US consumer goods company Procter & Gamble.
In 2015, the US Department of Justice laid 22 charges of fraud and market manipulation against trader Navinder Singh Sarao. As the case played out, we were introduced to high-frequency trading terms, like spoofing, layering and front-running.
Layering is where a high-frequency trader places multiple bogus orders, which are quickly cancelled to fool the market into believing a big price move is underway. Spoofing is a similar (and equally illegal) tactic where the trader places hundreds or thousands of orders for an asset to trick other traders into jumping on the bandwagon to drive prices either up or down. The spoofer has no intention of executing on these fake trades.
Front-running is where a trader has advance knowledge of information that might move an asset price. Front-running can occur when a broker receives a large order to acquire stocks on behalf of an institutional buyer. The order is large enough to move the market price, so the broker can front-run it by placing an order of his own before executing the large institutional order, knowing he can make a quick profit from the subsequent movement in price. Front-running, a close cousin to insider trading, is illegal in most markets. HFTs have been accused of all these transgressions, and more.
Research by Markus Baldauf and Joshua Mollner at the Kellogg School of Management at Northwestern University points out that only big hedge funds or investment banks can afford the heavy technological cost of entering this market. They conclude that HFTs have reduced the ‘spread’ costs of trading (the difference between the price of buying and selling an asset) and are therefore a benefit to the broader market. On the downside, HFTs were found to amplify market moves already in play, and that can distort markets. Their recommended solution is to introduce slight delays in the release of market data to avoid this kind of market arbitrage, or profiteering by jumping onto a stock slightly ahead of the pack with a view to making a profit at the expense of the laggards.
“HTF accounts for a growing percentage of total JSE trade, perhaps 35% to 40% of the total.”
Tshepo Maseko, Legae Peresec
Rules? What rules?
There have been attempts to regulate HFT in Europe and the US, but new rules have done little to curtail the tsunami of trade from these machine-generated trades.
Tshepo Maseko, head of trading technology at financial securities firm Legae Peresec, says high-frequency trading in SA isn’t yet as sophisticated as in the US, but it’s getting there. “HTF accounts for a growing percentage of total JSE trade, perhaps 35% to 40% of the total. Most of the HFTs are in the most liquid top 40 stocks. Market participants use a variety of different algorithms to profit under certain market conditions, although a lot of the strategies are linked to the news cycle, since this is the main driver of price changes.”
There are thousands of algorithmic strategies. For example, an algorithm might trigger a buy on tech stocks when they drop to a predetermined price level, knowing this is where other traders generally buy.
Fanie Harmse, operations director for Swiss-based FX & Project Management, a crypto and financial trading tools company, says HFT has made intra-day trading less consistent than was the case in the past. He refutes the suggestion that HFTs have much influence over long-range market trends.
“High frequency algorithms can’t influence prices in a sustained matter. Their goal is to get in and out as quickly as possible, making a tiny gain on each trade. They benefit through extremely fast execution that normal traders don’t have access to. In the longer term, price movement is largely controlled by big international banks, governments, corporations and hedge funds. These big players aren’t interested in intraday noise, even though they will often have systems to benefit from it through market-making. For political, economic and other reasons, they’re more interested in longer-term trends.”
Muddying the waters
Harmse says the widespread use of algorithms and artificial intelligence has reduced any competitive advantage, making it harder to profit from market moves that happen in milliseconds. That’s a view born out by international research – HFT profits have dwindled over the last decade. FX & Project Management has developed an algorithm called Access that picks up only major market moves, which occur roughly 20% of the time. “The reason most systems fail is that they’re over-active in the market, and we know that the market is essentially moving sideways 80% of the time. It’s difficult to consistently make money in a sideways market. Our systems are designed to identify trends and ride these until the trend is over. I think it’s going to be much more difficult going forward for high-frequency traders to make the kind of money they made in the past, so intelligent algorithms are going to have to find new ways to identify profit-making opportunities.”
High-frequency traders have rejected claims that they’re a threat to financial markets, or that we’re sitting on a financial Hiroshima waiting for another trigger, as happened in the ‘flash crash’ of 2010.
Most of the HFTs are privately-owned companies and therefore don’t share their secrets, but by their own words they are a force for good. They provide liquidity when needed. Many of these HFTs are market makers, meaning they will both buy and sell assets, as and when needed. They make money on the ‘spread’, or the difference between the buy and sell price.
David Scholtz is a full-time private trader. He says: “HFT is definitely a growing phenomenon all around the world, but I don’t see it as necessarily evil, as is often claimed.”
Much of the volume generated from HFT shops is benign – providing liquidity when needed, or for portfolio rebalancing. For example, if there’s demand for Anglo American shares and very few people are selling, market makers will step in with stock on hand to sell. For that, they will want to collect a fee. Without this liquidity, the shortage of stock could drive stock prices higher than would otherwise be the case. In this sense, HFTs smooth the bumps in the road and have made it easier to acquire stock. High-frequency traders say the reason the ‘flash crash’ corrected so quickly was because algorithms picked up the market anomaly and reversed it.
The more egregious aspects of high frequency trading, such as spoofing and front running, have been banned, though will never be completely eliminated.
One thing is for sure: the machines are taking over the world of investment, and that’s not going to change.
