Modernizing the Marketplace: Artificial Intelligence in Trading
http://www.bloomberg.com/news/articles/2016-10-13/this-bank-beating-trading-powerhouse-doesn-t-use-human-traders
The rapid switch to artificial intelligence is a common theme nowadays, with companies seeking to make their job easier—and perhaps less foolproof—by using the purely computational mind of a computer. In the midst of this modernization is a trading company – XTX. A recent article in Bloomberg reports that XTX Markets, Ltd, a growing power in the foreign currency exchange market (reaching a position in the top 5 earlier this year), has been conducting trading purely off of mathematical models and data. The removal of the “human factor” presents interesting implications for the trading networks we have analyzed in class.
In class, we have worked with a balance of algorithm and human analysis. In trade networks, we used algorithms to determine when traders would trade—and what their prices and values might be. Underlying this algorithm was a concept of human desire—humans wish to make the best deals possible, in which they profit the most. In working with network power we saw human intervention within algorithms to an even greater extent; a human will almost never accept a bargain in which they are given 0% of the value, and this understanding shifted our expected values for different nodes. Human nature seems to be an inevitable factor in bargaining and trading—how, then, does XTX make it work?
First, XTX does, in fact, employ humans to do work—just not in the conventional sense of trading. Mark Spanbroek, vice chairman of the FIA European Principal Traders Association describes the work of these nonconventional traders: “It’s like autopilot…the pilots are sitting there to navigate it through bad weather”. Recently, XTX utilized human intervention to halt trading when the value of British Currency fell outside of acceptable range. Beyond small amounts of human intervention, XTX implements unique decisions within its algorithms—like making use of market “speed bumps”—ways of delaying certain kinds of trading. XTX’s algorithms, it seems, are so advanced as to achieve the goals within human nature without the traditional level of intervention.
Supercomputer trading isn’t yet perfect—it exhibits bugs and makes modelling mistakes just as humans do. With the advancement in artificial intelligence, however, one might expect to see a rise in increasingly sophisticated algorithmic trading companies—accompanied by a demand for the brilliant technical minds to build them.