Support Board

The LPPL model correlates the price with the time in the following formula:

log(p(t))=A+Btm+Ctmcos(ω⋅log(t)−ϕ)

In this formula, t is the time (measured backwards in days from tc), A is the logarithm of the price at tc, ω is the angular log-frequency, B and m measure super-exponential acceleration (B>0 and 0<m<1), C is the magnitude of the log-periodic oscillations and ϕ is the phase shift. (see: first article) The main difficulty in fitting this model is its high non-linearity. One strategy would be to try a sufficient space of combinations for the non-linear parameters (m, ω and ϕ), after which use a linear-fitting method to find the A, B, C parameters (e.g. using R).