As algorithmic trading methods multiply, a new study by Quantitative Services Group LLC, Naperville, Ill., has found that automated trading models that parse out trades at regular intervals can be front run by traders, driving up costs.
"The implication is be careful of what you're using and be aware that there are people out there who are looking" for ways to "sniff out" algorithmic trading strategies and take advantage of them, said John Wightkin, managing partner at QSG. "If they're market makers, they're going to back away and not necessarily provide liquidity, so liquidity charges go up, and people sniffing will get in front of the trade and drive up the timing risk around your trade."
The firm examined more than $1 billion in stock transactions from a large, unnamed institutional investor, covering more than 120,000 stock executions. Two algorithms were analyzed — one was a commonly available algorithm designed to deliver the volume weighted average price for the trading period, and the other was a custom algorithm that used a randomization technique in distributing orders to achieve the same goal.
The first algorithm generated 26 basis points in trading costs — a combination of liquidity charges and timing costs — compared with 2 basis points for the second method.