When you think of a trading floor, you think of screaming men in shirts; with notepads in hand and phones clamped between shoulder and ear. But those taps are long gone. They have been replaced by economists and mathematicians competing for the best algorithm.
“We have no sense, the computer does all the work,” says Dennis Dijkstra, co-CEO of Dutch Flow Traders. This can be seen in the workplace. Forty percent of its employees are software engineers. Yesterday’s financial specialist has only one place to play as a controller. “You need to build enough checks and balances and always keep checking it.”
More transparent and fairer?
BlackRock is also taking steps in that direction. According to The New York Times, seven out of 53 ‘stock selectors’ must leave their function, for the sake of mathematical formulas. 36 employees leave the company completely. With this move, the company leaves $ 30 billion in investments to computers.
Dijkstra sees the demise of the old-fashioned stockbroker as a step towards a fairer system. “It’s becoming more transparent.” Previously, only the men in the trading post had the latest information. Now that information can be found everywhere and with good systems available in seconds. A company like Flow Traders bases its systems on the amount of data. According to Dijkstra, being able to decide quickly based on more data gives a better price. In addition, trade is getting cheaper because expensive people have been removed from the process.
Yet there is a lot of skepticism. Algorithms have paved the way for high-speed trading: As soon as a large investor is seen buying large volumes of stocks, traders jump in between them to take advantage.
Here’s how it works: Large investors, such as pension funds, often buy such large amounts of shares that it can disrupt the market. Therefore, they never buy large quantities at once, but in several tranches.
Bart Schouw from Software AG, a provider of algorithm-based trading systems, explains how it works from London: “In the past, such an investor would just call and ask what a hundred shares in Shell cost. Immediately bought another hundred. And again and again. Until enough was bought. ” Now computers do.
How does flash trading work?
By smartly monitoring when a large investor is making such purchases, and analyzing that pattern, you can take advantage of a slightly faster connection. Just before the next tranche is bought by one computer, you quickly buy these shares and then offer them more expensive. Although only a few cents per can be obtained. share, large volumes make this business very lucrative.
This is legal but is considered unfair by many. “One has to imagine standing in line at a counter to buy Shell shares,” Schouw explains. “But just before you want to pay, someone jumps in and quickly buys those shares. Then he turns around and sells them to you just a little more expensive.”
By moving into offices close to the stock exchange building and buying computers with higher computing power, just the extra millisecond that gives a profit can be achieved.
“Even though the battle for speed has been won,” says Schouw. “It’s not going to get any faster than this. The battle that is now raging is primarily about complexity.” Large investors need complex constructions that an algorithm does not recognize. This can be done by making large orders, such as driving pension funds through the market, as unrecognizable as possible.
Dijkstra rejects resolute criticism of algorithm trading. “This is something of all time. If a pension fund sells a few million in Shell shares, it is simply changing the market. It is supply and demand. In the past, traders have also benefited from that information.”
According to Dijkstra, trading becomes easier by sitting between transactions. Because different exchanges are connected to each other and have data sets that are as complete as possible, according to him, a more fair price is created. Supply and demand are better brought together in this way.
And what about people?
According to Albert Menkveld, professor of finance at the Free University, advocates have made a point. “With very large orders, you see that flash dealers fill their role really well in the beginning. They then sell what it takes. But after a while, they start buying and sell it so quickly on, then you see that f. for example, a pension fund pays way too much. “
That big companies like BlackRock are slowly replacing part of their arsenal of financial experts with algorithms does not surprise anyone in the market. Opinions, however, are divided on where this trend will end. According to the experts, the traditional dealer will have to keep checking the algorithm until further notice. “The responsibility always lies with people: you can’t put a machine in jail if it goes wrong,” says Schouw.
It is not yet possible to implement artificial intelligence that will be fully self-governing. “As a financial institution, you need to be able to explain why you have made decisions, with artificial intelligence doing it for you, it’s not possible. You always have to understand it yourself.”
In addition, we are simply afraid of it, says Schouw. “Institutions should not think that such a system is going crazy because of an error in the code.”
According to Professor Menkveld, a much bigger problem lurks in the near future: When algorithms start making decisions, only a small part of the information will be taken into account. “Hard information weighs a lot, while ‘soft information’, such as a change of board, is poorly interpreted.”
According to Menkveld, this will significantly affect the functioning of the stock exchange: estimating the value of companies. “If it is no longer done correctly, it weakens the economy. Companies that should raise money through an issue can miss it, while bad companies continue to exist.”