The Future of Trading
10 Top CIOs on Wall Street | Why Mutation Will Matter | Buyside Traders: Not Taking Orders Any More | Calculating Cost First, Trading Second | Trading Desk, in One Box | Changing the Minds of Star Programmers | 2015: The Always On Desk
Why Mutation Will Matter
November 4, 2010
Hyde Park Global Investments in Atlanta uses no analysts, portfolio managers or traders. Its operations are entirely based on robotic trading, using artificial intelligence and what it calls “genetic” algorithms to execute trading instructions.
Adam Afshar is its president and chief executive officer. Afshar is a former managing director at Bear Stearns, with an MBA from the University of Chicago. Securities Technology Monitor emailed Afshar to find out why he trusts machines more than men and women, once trading begins.
1. What do you mean by ‘100% robotic trading’?
It means 100% of Hyde Park Global’s operation from idea origination to execution is machine based and we have absolutely no discretionary human intervention. The humans are involved with coming up with the hypothesis, testing it, modeling it and building the artificial intelligence around it.
2. How are you using artificial intelligence in this?
Computers get the data on over 4,000 stocks and analyze this data by making more than 20,000,000 calculations per second.
We do not use computers and robotic trading because it is cheaper or more efficient. We use it because it is demonstrably better and maybe the only way to do what we are doing. It gets us better returns. And we make fewer “mistakes” in judgment.
4. Why has artificial intelligence, after all these years, not been widely adopted on Wall Street in its trading programs?
Let me answer you with questions of my own. Why after all these years do people still smoke? How hard do you think it is for folks just to change their sitting spot at the dinner table? Have you ever tried to sleep on a different side of the bed? Change is hard. Note that most innovative financial firms are young or new firms that have brought in a fresh scientific and engineering multidisciplinary approach to investment.
5. How do your genetic algorithms differ from other algorithms?
Our genetic algorithms have the ability to recognize complex trading patterns from terabytes of historical data. They construct and optimize automated trade decision logic from these patterns. In contrast to many other algorithms, such as neural networks, our trading models are always understandable – this is what I would call a clear crystal box, not a black box.
6. Why do you call them ‘genetic’?
It is an evolutionary optimization process, which uses the principles of natural selection and "survival of the fittest" to solve complex problems too large to deal with in an exhaustive manner.
In our case, we start by constructing a set (or population) of many random variations (or chromosomes) of the trading strategy we wish to optimize. Each variation is traded over the learning period and ranked according to a defined objective function, which may be a Sharpe ratio or Omega ratio.