By Murat Aslan
ISTANBUL; 03 Feb 2019; AA: “At the end of the World War II, the average holding period for a stock was 4 years. By 2000, it was 8 months. By 2008, it was 2 months. And by 2011 it was 22 seconds, at least according to Professor Michael Hudson’s estimates from University of Missouri-Kansas City,” says Scott Patterson in his groundbreaking book, Dark Pools.
Since computerization started to dig into the daily life of human beings in the 1950s, there was no going back to old times. Progress was shockingly fast, and with computers facilitating things more and more for intelligent people, they spotted greater arbitrage opportunities (‘arbitrage’ here meaning the probability of making money in the markets with no risk of losing). A great many PhD holders from fields such as mathematics, physics, chemistry or engineering flooded into High-Frequency Trading (HFT) firms to create algorithms that benefit from any irregularity in the markets.
A firm is considered to be HFT if its daily volume of share trading amounts to thousands, if it is using extraordinarily fast computer programs and is located close to the market-makers to eliminate delays, can end the trading day as close to zero inventory as possible and do all the trades in an extremely short space of time.
In the era of conventional trading, humans used to execute the buy or sell orders on the trading floors of markets, and their capacity was naturally limited. Their human limitations gave computers the edge over them since these machines were able to process things more quickly than any human being ever could. Quants, that is, mathematicians who use quantitative techniques to predict market moves, started coding algorithms with complex designs that enabled firms to profit out of thin air in seconds.
The revolution of the machines in trading knocked the door around the 1980s and those smart enough to fathom what was happening welcomed the opportunity wholeheartedly. Exchange markets such as NASDAQ or NYSE became digitized over time and, in so doing, put all investors on an equal footing, in a way, in terms of order of priority, thereby removing a major obstacle in the way of computerized systems’ development. Thus computerized trading well and truly got underway to eventually achieve incredible speeds and execute orders in nanoseconds (a billionth of a second), according to Maureen O’Hara from Cornell University.
Speed: a must in trading
Data was a rich seam to mine for the traders but when it was processed manually, it tended to become obsolete by the time the operation was over. Handling the available data by algorithms meant exploiting the market before anyone noticed and that meant huge amounts of money. So, speed increasingly became a sine qua non for HFTs. Liquidity, on the other hand, might be considered as the beating heart of the markets. One of the advantages of HFTs was that they liquidated every market they operated in.
Even though liquidity is a multi-dimensional concept, one of the easiest ways to measure is to look at the bid-ask spread. A bid is the maximum price that a buyer is willing to pay and an ask is the minimum price a seller accepts for a security. The difference between the bid and ask price is called a ‘spread’. The narrower the spread, the more liquid the market is. If there is no buyer in the market at the time when a seller is asking, the market-maker might buy and put the security in his inventory with a decrease in the risk that he is bearing. Since quotes change rapidly, the market-maker would not want to lose in the trade, either. Thus, the seller loses a portion of the price due to illiquidity.
Bid and Ask spreads created an arbitrage opportunity for HFT firms as, for example, they could buy a thousand shares of Apple for $50 and sell them for $50.125 in seconds in different markets, and even in the same market sometimes. And they could do this over and over again until the price of an Apple share was $50.125 in each market. According to a study conducted in Columbia University, the minimum spread has gone down to $0.01, whereas it used to be $0.125 in 1997, for instance.
Quants had found a gold mine in the markets with the creation of algorithms, which analyzed the data and executed the order with speeds that human beings could never reach. A quote was a quote. Anyone who was behind the market was ripped off, even the market-makers themselves.
The faster, the better
Computer programs enabled HFT firms to offload paperwork and watch several -- later many -- stocks, at a time, which created another edge over the conventional traders. Capturing imbalances between quotes thus, able to instantly sense opportunities, and equipped with lightening-speed, algorithms kept profiting. “The faster, the better” was the motto of all HFTs.
In the 1990s, markets were full of arbitrage opportunities for those who were handling their trades through intelligent computer programs. Some coders realized that they could even bypass NASDAQ or NYSE, which charged $2.5 per trade. So, “the Island” was established by tech-genius Joshua Levine, charging $1 per trade. Island was an electronic trading pool where people with stocks could dive and trade.
By late 1996, almost half of the NASDAQ’s automated orders came from the Island: 5.6 billion shares changed hands, equaling some $22 billion. On April 4, 2001, the Island conducted 126 million exchanges of shares, valued at over $18 billion. Nearly 15 percent of the trades of NASDAQ were coming from the Island as of 2000, according to Patterson. A study in 2001 showed that 25 percent of the trades handled through the Island were done under two seconds. Patterson mentions that an electronic trading pool called BATS Chi-X Europe handled more than 25 percent of all stocks trading in Europe in 2011.
Price beats time
The execution of orders was made on the basis of ‘first come, first served’. Indeed, HFT firms coded many algorithms to beat their competitors in the race but the rule was still valid. However, HFTs were not bound by the conventional limit order, an order type that ordinary people are offered by institutions. They had also discovered anomalies in the market structure and modified their orders accordingly. They had learned that price beat time.
For example, if a share was being sold at $50 in the market, they realized those who offered $50 would be queued and whoever was willing to pay $51 would take the lead and buy the stock still at $50. According to a rule set by US stock exchanges, if a buyer would offer a price 20 percent higher than the market price, they would be the first in line to buy. Of course, coders added this valuable information into their algorithms as soon as they found out about it.
According to a US Securities and Exchange Commission (SEC) report released in 2014, one of the characteristics of the HFT was, they were co-located around the markets to minimize delays in the delivery of orders. Firms built huge computer labs or moved present ones near electronic trading pools such as the Island to grab a favorable position in order queues.
One of the fastest firms, based in Kansas, observed that even though it was considerably faster than others, its orders were getting late due to their distance to the pools. So, they decided to move closer and connect to the server itself via cables. The improvement was dramatic. They were now able to execute 20 trades in one-fifth of a second, whereas this used to be the time a single order took to travel from Kansas to New York.
Since not every firm was able to move near electronic pools, a new race was launched among tech companies to produce the fastest fiber-optic cable. Two companies started out a half-billion-dollar transatlantic fiber-optic cable project, connecting London and New York. To protect the cable from sharks or ships, they buried it 1.8 meters under the seabed throughout the Atlantic Ocean. However, even fiber-optic cables were soon to be abandoned. Microwaves could transmit the orders 3 milliseconds, one millionth of a second, faster than cables.
According to Menkveld, as he puts in his 2016 study on nanosecond data, 20 percent of the trades were concluded in a millisecond, that is, a millionth of a second. Hu, Pan & Wang (2014) suggest that the algorithms are so powerful that they could reach the official macro announcements two seconds in advance, just enough time for an Artificial Intelligence (AI) to decide and act. In other words, HFTs had themselves positioned even before an announcement was made.
The transformation to advanced computer-based systems was inevitable in view of AI-guided algorithms. Thus, many markets from the East to the West, including Turkey’s Istanbul Stock Exchange (later BIST), replaced the old systems with the latest technology around 2010.
Is an ‘AI-Guided Warren Buffet’ possible?
Could AI be designed to make a Warren Buffet, the billionaire investor and CEO of the Berkshire Hathaway Company, out of the HTFs? Could a computer program follow the news, pull the data, analyze it and decide which security to buy in seconds? According to Patterson, in 2011 there was such a program called Star, coded by an intelligent mathematician Spencer Greenberg. The program had better returns than S&P 500 in 2007 with 17 percent overall against the 5-percent gain of S&P 500. Star kept up this performance for your years in a row until 2011, using only AI coded by a math genius.
Algorithms are beyond denial and machines could become the only operators in trading in the future. This might lead ordinary investors to withdraw from the markets, where they trade mostly for the purpose of long-term benefit. In addition, permitting machines that are coded only for a single purpose might harm the firms that go public to fund themselves in the first place. In order to have a healthy market environment, there should be a regulatory basis to protect the firms as well as the investors.
