This dissertation tackles the problem of non-normality in the distribution of returns and attempts to formulate a proprietary trading strategy to arbitrage the markets using appropriate statistical and mathematical tools. The first essay provides fundamental understanding to fractional Brownian motion (fBm) process, its characteristic Hurst exponent, and the concept of unit root in time series data. The study shows that a simple autoregressive (AR) process with suitable lag coefficients is able to effectively replicate the fractal time series and preserves its characteristic Hurst exponent. More interestingly, an equation that defines the relationship between the AR lag coefficients and the Hurst exponent that described a particular fBm process is also derived. The second essay introduces the concept of excursion measures and illustrates how the Itô‟s excursion theory can be used as a tool to understanding fractals. The excursionsvalued process is shown to follow a binomial distribution which is a robust substitute for Poisson distribution as suggested from the theory. The results also show that a process with low Hurst exponent or short-memory process has higher mean excursion measure at low excursion length as compared to a process with high Hurst exponent or a longmemory process. On the other hand, we see systematic wandering with longer excursion in a long-memory process with Hurst exponent higher than 0.5. Based on the discovery from the first two essays, the third essay combines these findings together to form a trading strategy called "Hurst Trading" with trading signals generated from the fluctuation in the dynamics of the Hurst exponent across time, among other indicators. We find that for the period between 2002 to 2011 the Hurst Trading strategy is able to outperform the traditional momentum strategy and the "Buy and Hold" strategy by a wide margin on stock trading in the DJIA Index, SPX Index, and R2500 Index. Furthermore, the more fractal the process is, the higher the chance that the Hurst Trading algorithm would be able to correctly time the entry/exit points in the market.