Portfolio selection algorithm under financial crisis: a case study with Bursa Malaysia

Abstract

Most study on online portfolio selection algorithm focus on the theoretical derivation of optimal regret bound or empirically validates portfolio cumulative return and its variability. This study investigates the behavior of algorithm under financial crisis based on 2008 stock trading in Bursa Malaysia, a market in small open economy whereby trading actions could not exert impact to the spillover trends from US and Europe. The equity returns data generating process under this scenario is an AR process with positive lag as such algorithms arbitrate between relative growths like Anticorrelation, constant rebalancing are not performing. Whereas algorithms that search for optimal portfolio at each transaction such as Universal Portfolio, Convex Optimization approaches are able to reverse the downward trends of portfolio before market recovery, dampen downside variability and deliver lower extreme returns. We also explained the expendability and practicality of the Convex Optimization approach for future development of automated trading scheme.