It has been well known since the pioneering work of Mandelbrot
that market price fluctuations are poorly modeled and described by
discrete time random walks or continuous diffusions driven by stan-
dard Brownian motions. In this talk we analyze financial time series
such as oil or bitcoin prices. The data exhibit very interesting multis-
cale correlation structures that can be characterized by a time-varying
volatility and Hurst exponent and that can be used to identify regime
shifts for the market prices. This is a joint work with Knut Sølna (UC
Irvine).