Beating the high frequency traders

MarkletWatch:Here is a typical scenario to show how this works: An asset manager legally discovers better information of a stock than the market does and trades to exploit that information. After the order is filled, a “back-runner” legally observes the institution’s filled order. The back-runner could be an HFT that uses sophisticated pattern recognition software to determine the existence of a large investor trying to buy or sell. The back-runner then competes with the institution by using that order-flow information to its advantage.

In this situation, the HFT has no crystal ball; it cannot see an order before it reaches the market. Instead, the HFT is watching the market all the time looking for patterns that indicate the intention of a large investor, as that suggests that a particular stock is over or undervalued. When the HFT sniffs out a large investor, it can become a competitor with its own transactions, driving the stock price up or down.

The best response of institutional investors is to introduce some “noise,” or the appearance of randomness, to cover their tracks. For example, if the real goal is to buy 100,000 shares, then the investor could include some sells in the mix of transactions to essentially play hide-and-seek with the HFT and mask its true intent. It also could change its buying pattern so that the number of shares per trade and the timing of the trades appear random. This use of randomization makes it riskier for the back-running HFT, as it can’t be certain of the institutional investor’s real plans or even its presence.

I think we knew about this solution for five years, at least since I mentioned it.  This is straight out pf Shannon, first measure the uncertainty level of the market, then trade just outside of that band, breaking up your trade as necessary.   The sad point is how clueless traders are regarding statistics.