3 Incorrect Notions About Statistical Arbitrage

Share on Facebook0Tweet about this on Twitter0Share on LinkedIn1Share on Google+0

Statistical Arbitrage is quite popular among HFT and quant traders. Any arbitrate opportunity arising due to misquote in prices can be very advantageous to a technological driven trading strategy. Such opportunities may exist only for a few seconds in the market before the prices get re-adjusted, but those few moments are enough for a speed trading strategy to make money from.  However, there are many misconceptions about the statistical arbitrage strategies in the industry. We point out a few here.


Statistical Arbitrage is (NOT) risk free

Arbitrage strategies are considered risk free opportunities where you buy in one market and sell in another market as long as the price difference exists. For instance, if say gold is cheaper in Dubai than in London and you buy it in Dubai and sell it in London to make a profit from the difference; you had an arbitrage. However, in case of statistical arbitrage it is not so simple.

Statistical arbitrage uses mathematical models to identify statistical mispricing in the prices of highly correlated stocks. The arbitrage opportunity depends on the ability of the market prices to return to the predicted means. In periods of financial crisis, such strategies would be quite risky.

Pair Trading and Statistical Arbitrage are (NOT) different

Both have overlapping meanings which involve finding highly correlated pairs of stocks which are mispriced due to market inefficiencies and can be leveraged to make profits until the prices revert back to the historical or predicted values. The underlying concept is of mean reversion.

The only difference between the two can be in terms of number of trades, speed and execution type manually or automated.

Correlation and Co-integration are (NOT) same

High correlation does not necessarily imply high co-integration. Correlation reflects the relationship between two stocks in terms of price returns . However, it is not a robust indicator that measures long term movements because it is sensitive to small time deviations. Co-integration measures the relationship between price movements of two stocks over long time periods and can serve as an additional parameter to find the pairs.

Share on Facebook0Tweet about this on Twitter0Share on LinkedIn1Share on Google+0

Leave a Reply

Your email address will not be published. Required fields are marked *