Online portfolio selection

Online portfolio selection is an algorithm based trading strategy that seeks to optimize investment returns by repeatedly reallocating capital across a set of assets as new price information arrives, rather than computing a single static allocation in advance. Where classical portfolio theory balances expected return against risk to choose one fixed mix of holdings, online portfolio selection algorithms instead aim to maximize the cumulative wealth an investor accumulates over a sequence of trading periods, an objective closer in spirit to the Kelly criterion for long run growth than to mean variance optimization. Different algorithms in this family pursue that goal through different strategies: momentum based methods shift weight toward assets that have recently performed well, contrarian methods do the opposite by favoring recent losers on the expectation of reversion, and other variants draw on pattern recognition, fuzzy logic or explicit risk assessment to decide, period by period, how to redistribute the portfolio's capital.