Experimental results announced at the recent International Joint Conference on Artificial Intelligence in Barcelona show that robot trading agents in foreign exchange markets outperform humans. The experiment is a follow-up to the infamous 2004 IBM experiment when human traders competed, for the first time, against state-of-the-art computer trading agents – and lost. What is new about the 2011 results is the strategy that the robot traders used as a measure of comparison.
The latest experiments, run by Marco de Luca and Professor Dave Cliff of the University of Bristol, benchmarked the “adaptive-aggressiveness” (AA) strategy designed by Dr. Krishnan Vytelingum, Professor Dave Cliff and Professor Nick Jennings of University of Southampton in 2008 (see reference below). The “AA” strategy was shown to be the leading strategy able to beat both humans and robots following other strategies.
The foreign exchange market (also referred to as “Forex”) is a very interesting application area for robot traders and it is no accident that it has been chosen by researchers. The market was created in the 1970s following the relaxation of the “Bretton Woods system” of international monetary management set up after World War II. It allows banks, commercial organizations and individuals to perform international transactions across various currencies as well as trade currencies. The gradual deregulation of money markets has meant that nowadays Forex has a daily turnover of around 4 trillion dollars. Forex is regarded as the closest market to the ideal of perfect competition. This is because of its various characteristics including the low profit margin compared to the huge trading volume, its round-the-clock (except weekends) operation, as well as the variety of factors affecting the exchange rates.
Market psychology, as always, plays a dominant part in effective bidding. Unsettling international events – such as the euro-crisis or the debate over raising the US debt ceiling – create herd-like “flights to quality” in haven currencies, such as the Swiss Franc witnessed lately. The cognitive systems of human traders are at a disadvantage here. A psychological bias known as “anchoring” makes investors focus too much on the relevance of outside events to currency prices. Robots don’t. Bereft of cognitive biases they learn, adapt to dynamic market fluctuations and make efficient, and profitable, bids. They truly “buy the rumor and sell the fact”. As a result of their success 70% of trading in Forex goes through robot traders.
There are many issues raised when confronted with automatic trading in financial markets. As always, the most fundamental ones are political rather than economic. What is the purpose of a market? According to the anarchic capitalism of the Austrian school the purpose of a market is to maximize its value, full stop. Freeing markets from regulation increases competition, optimizes prices and spurs innovation, economic results supposed to increase wealth and ultimately make this world a better place. In this context human as well as robot traders have a clear goal: compete on equal terms and increase profits, a goal served by various trading strategies including AA.
Nevertheless, anarchic markets appear to be in conflict with much of the utopian narrative of robots. In this narrative there is a strong humanistic – or “post-humanistic” – goal towards a perfect society of social justice where humans and machines co-exist in productive harmony. Arguably, this is a quasi-socialist ideology where machines regulate society and the economy, rather than deregulate. Current trends in applied robotic capitalism push towards a more dystopian version of the future, where robots beat humans in their game but the result is not necessarily a more “just” world. In fact “justice”, in whatever sense, is irrelevant.
References: Vytelingum, P., Cliff, D.,Jennings, N (2008). Strategic bidding in continuous double auctions. AI