What the Data Can't Tell You: Plucking the string of the market

Image via Wikipedia
So much time and effort has been placed in the pursuit of analyzing past stock market data sets to mine for statistically significant anomalies. This is a great process for determining what would have made you money in the past, but is lousy for predicting how the strategy will perform when large amounts of money are employed to capture the discovered anomaly. Oftentimes, especially in statistical arbitrage, the presence of one or two more large players can tip a once profitable strategy into a money pit.

The key to arbitrage is being the first or second to find it... and staying small. Scaling up a strategy is hard, but making small amounts of sure money is easy for those patient enough. Perfecting a stable of small but consistently profitable arbitrage strategies is the key to long term income.

A successful arbitrageur should be able to be both statistician and wildcatter. We must pluck the string of whatever market instrument we have assembled and listen to what sound results. We must model the interaction of our strategies with the market as judiciously as we derive the theoretical arbitrage itself. Otherwise, taking a strategy from theory to practice is impractical and useless.

This financial crisis has taught that even the most statistically inclined of us must have a sense of just going with the flow. Make a decision. See what happens. Go from there. Use the tools that have worked before but with new assumptions.