Non-random behavior in financial markets

What is the nature of the price formation process? This thesis’ project uses an ontological approach to analyze the time series of historical prices in financial markets, at different frequencies, and across multiple asset classes, in order to understand (i) the existence of any non-random pattern, (ii) the eventual information that they may contain, and (iii) what may be their implication at microstructural level. In the first (i) part we developed a pattern recognition algorithm to extract consecutive trend lines from prices, obtaining a robust statistical evidence of a significant and systematic memory in historical prices. In the second (ii) part we implemented an asset allocation model to test the value of the information contained in the memory of the process. The investment strategy significantly overperformed the market for every security and frequency analyzed. In the third (iii) part we studied the effect of the deterministic patterns at microstructural level demonstrating how their presence biases the well-documented relationship between trading frequency and volatility. Our results, in fact, are partially in contrast with the main literature, highlighting a non-stable relationship through the trading day.