This document presents an improved messy genetic algorithm (mGA) called fast messy GA (fmGA) that aims to optimize difficult problems more rapidly and accurately. It addresses the complexity issue of mGAs by introducing two key improvements: probabilistic initialization and building block filtering. Through experiments on concatenations of deceptive k-trap functions, the fmGA showed it could solve such problems with overall complexity of O(l log l), demonstrating faster processing than standard mGAs while maintaining their theoretical advantages.