Thursday, April 22, 2010

Open systems, financial data, and algorithmic trading

Economic data suffers from a peculiar conundrum. Most of this data is collected in the form of asset prices at different points in time. Unfortunately, since time is embedded, the points of data cannot be described as neat fungible packets such as what we find in physics or chemistry...these fields have data that operates independently of historical human events and meddling.

This is a somewhat convoluted way of saying that it is difficult to distill historical social data without reference to the unique period of time from which that data is derived. This is why finance and investing is more art than science.

Unless, of course, you are an algorithmic trader. These quants scalp spreads and collect exchange rebates with efficiency that would alarm most casual observers. The art of parsing history and data is irrelevant to these traders, which makes them extremely attractive as investments given the expectation their returns would be uncorrelated with market fluctuations.

However, the profits that were there for easy taking in my opinion are eroding fast in this field. A flood of competition, trade secret expropriation (read: some guy steals your code and uses it against you) is producing the usual effects, and the "asset class" that is algorithmic trading is experiencing its own bubble which will end (through the abruptness of the guillotine via government regulation or by more mundane natural causes) soon.

No comments: