Systematic trading ain’t what it used to be

Blog first appeared in DanCollinsReport on May 1, 2013

I noted earlier how it was more than a little odd for the market to react to one bad bit of news when it came from a tweet hacked from the AP Twitter account. Some analysts have speculated that automated algorithms that scan headlines may have been responsible for the quick 140-point drop in the Dow Jones Industrial Average on April 23.

CNBC’s Patti Domm reported that Kenny Polcari of O'Neill Securities stated, "That goes to show you how algorithms read headlines and create these automatic orders – you don't even have time to react as a human being. I'd imagine the SEC's going to look into how this happens. It's not about banning computers, but it's about protection and securing our markets."

To follow-up on what Futures Industry Association President and former Commodity Futures Trading Commission Acting Chairman Walt Lukken referred to as the “hash crash”, let’s take a look at the notion of trading off of news headlines and tweets vs. an expert algorithmic trader.

I am not sure what exactly constitutes an entry though there are obviously some traders who traded off of a single input, the erroneous hacked AP tweet. Others suggested algorithms that kept a score of positive and negative headlines with some mathematical equation used to trigger a long or short trade based on the percentage of positive vs. negative headlines/tweets.

This seems to be folly.

I interviewed legendary trader Bill Eckhardt a few years ago and he talked about his systematic approach to system building. He refers to the various elements within a trading system as degrees of freedom. “Every time you need a number to define the system, like a certain number of days back, a certain distance in price, a certain threshold, anything like that is a degree of freedom,” Eckhardt said.

Eckhardt was talking about his attempt to avoid curve fitting, or over-fitting as he put it. The key for Eckhardt was to have as many samples of how a trading rule performed in the past. For it to even be considered for his system he required 1,800 instances of a signal and preferred more. “Typically we would have 15,000 trades of a certain kind before we would make an inference as to whether we want to do it,” he said.

comments powered by Disqus