The term statistical arbitrage refers to the practice of using sophisticated mathematical models to identify potential profit opportunities from a pricing inefficiency that exists between two or more securities. Statistical arbitrage requires the use of high speed computers, computational models, as well as complex trading systems.
While stock exchanges are considered efficient markets, there are instances when the mispricing of one or more securities provides the opportunity for profits through statistical arbitrage. The approach requires the trader to use both quantitative models and datamining techniques. As is the case with high-frequency trading and quantitative trading, the costs associated with statistical arbitrage are high. For that reason, the technique is typically limited to hedge funds and large investment banks.
Also known as pair trading, statistical arbitrage involves the buying and selling of securities based on their long term relationship to each other. That is to say, the price of stock for companies in the same sector of the economy may follow each other in some statistically significant manner. When that relationship varies from what is considered “normal” in the near term, an arbitrage opportunity may exist, since the expectation is the long-term relationship will eventually return. For example, there may be a long-term relationship between the price of FedEx common stock and that of UPS. When that relationship varies from the norm in the near term, price arbitrage exists if that relationship eventually returns to normal.