Why will 2017 be a better year for statistical arbitrage? The answer lies in the fact that the dispersion of weekly stock returns, basically the gap between the top and bottom percentile, has reached the highest level since the financial crisis in 2008/2009.
Asset classes, in general, have seen past correlations disappear. Due to the resulting increased dispersion, opportunities for statistical arbitrage have increased substantially. Asset prices that typically have moved in near lockstep like US equity and oil have become less correlated, which portrays a challenge to portfolio diversification.
Uncertainty arising from the political risk in the US and Europe, with the recent election of Donald Trump, and the upcoming elections in the Netherlands, France, and Germany, certainly adds to the potential to shake up markets and create dispersion.
Statistical Arbitrage uses quantitative and automated trading systems that seek out temporary and small misalignments in prices among securities. The most popular method is pairs trading. Stocks are paired up based on an analysis of fundamental similarities or the beta.
By pairing up similar companies whose returns are highly correlated, but one being priced more aggressively, profit can be made. Quantitative hedge funds like to use this strategy nowadays to deliver alpha for their clients and outperform their benchmarks, although it was an investment bank that started the trend – Morgan Stanley’s proprietary trading desk in 1990 invented pairs trading.
Finding The Correlation
One can briefly elaborate this process from the simplified perspective of an algorithm. An algorithm will convert the time series of stocks into binary signals (0,1) and then try to find a similar stock with the same binary time series. If the correlation is detected between the two stocks binary sequence, the algorithm will start tracking both securities.
The algorithm will only take action when divergence starts taking place between the stocks. If the delta > x, then the algorithm will buy one stock and sell the other. This divergence can be only exploited with high-frequency trading underpinned by a ultra-low latency infrastructure, due to the small timeframes of the tracking taking place and the many pairs an algorithm has to
This divergence can be only exploited with high-frequency trading underpinned by a ultra-low latency infrastructure, due to the small timeframes of the tracking taking place and the many pairs an algorithm has to analyse. The Statistical arbitrage model is widely based on data mining, discovering patterns in historical pricing data to be exploited by the quantitative models. This can be used to forecast future time-series and give the systems more information input to what to look for in the tracking process. All the algorithm does is to invest funds into overbought or undersold spreads that arise from the divergence.
By taking relatively small positions and because there are so many of those opportunities, the law of averages would ensure almost statistical certainty. That is because the probability of profiting from those many value bets is very high and can generate substantial returns, depending on the capacity being allocated towards this investment approach.
Statistical arbitrage is really interesting as the binary sequence does not necessarily correlate two stocks from the same sector but more their pricing movements, that provide valuable insights into stock-picking and more opportunities to deliver alpha in an uncertain market environment ahead. Potentially the year ahead will teach more about newly correlated asset classes and show how one can see a correlation change. Exciting new times create exciting possibilities.