Man Versus Machine
With algorithmic trading, it seems like the evolution of electronic trading has taken a quantum leap. Not only can these quantitative models slice and dice large orders into smaller pieces, they are also intelligent enough to monitor a stock's trading pattern and fill orders in a way that masks their placement.
Today, human beings - financial engineers and traders - are the masterminds behind the algo-trading strategies. But what is the future of these so-called black-box trading strategies? Given where we've come from and how far we've gone, is it reasonable to suggest that automated-trading systems may someday replace the human trader altogether?
After all, these are not dumb algorithms. Traders and software experts on the brokerage side tell me that the algorithms are learning from the data they process. "All algorithms learn from the data they interpret," says Rob Flatley, managing director at Banc of America Securities. In fact, BofA acquired Vector Partners, which used algo trading to run a statistical arbitrage fund and had processed 10 billion shares through its algorithmic-trading servers. "The more data running through it, the smarter the algorithm can behave," Flatley says.
After monitoring voluminous amounts of data and spitting out orders, the algorithms learn and readjust themselves to changing market conditions. Say a trader tells the algorithm the order should participate in 15 percent of a stock's average daily trading volume; the algorithm could change that to 7 percent.
So what does the future hold for the trading profession?
Today, according to Dan Mathisson, global head of advanced execution services at CSFB, the trader sets the big-picture parameters once every few hours or once a day, while the computer responds to changes in the market, running updates as often as four times per second. But the trader is still in charge. In the next stage, algorithms will learn on their own and suggest mutations. "That is definitely down the road, and the math is going to get higher and higher power," he says. For now, Wall Street's algo engineers are focusing on replicating the way a really good trader trades. "That will flip to the point where the computers will trade better than any person can hope to trade," he asserts.
But don't worry. "We're not at the point yet in trading where the computer is beating the equivalent of Kasparov in chess," Mathisson says. That could take another 25 years, he adds. |
