In this project, we plan to study whether the roughness observed in realized volatility is a true feature of the underlying spot volatility or an effect of the estimation procedure. We will reproduce a (normalized) p--variation roughness estimator and test it on simulated Brownian, fractional Brownian, and stochastic-volatility models. We will then apply the method to market data and examine how the estimated roughness changes with sampling frequency, volatility window size, and block size. The goal is to build a simple and reproducible workflow for testing rough-volatility claims using both simulations and financial data.