The proposal of this paper is to provide a simple angular random walk model to build up polypeptide structures, which encompass properties of dihedral angles of folded proteins. From this model, structures will be built with lengths ranging from 125 up to 400 amino acids for the different fractions of secondary structure motifs, which dihedral angles were randomly chosen according to narrow Gaussian probability distributions. In order to measure the fractal dimension of proteins three different cases were analyzed. The first contained α-helix structures only, the second β-strands structures and the third a mix of α-helices and β-sheets. The behavior of proteins with α-helix motifs are more compacted than in other situations. The findings herein indicate that this model describes some structural properties of a protein and suggest that randomness is an essential ingredient but proteins are driven by narrow angular Gaussian probability distributions and not by randomwalk processes. The manner in which a protein folds from a random coil into a unique native state in a relatively short time is one of the fundamental puzzles of molecular biophysics. It is well accepted that a unique native three-dimensional structure, characteristic of each protein and determined by the sequence of its amino-acids sequence, dictates protein functions. The folding process should involve a very complex molecular recognition phenomenon depending on the interplay of many relatively weak non-bonded interactions. This would leads to a huge number of possible final conformations under conventional molecular optimization methods based on the search for the minima of the energy hypersurface. This number, which should increases with the number of the chain's degrees of freedom, however, is severely restricted during the real folding process, excluding relevant portions of the energy landscapes as far as an extended or random conformation is chosen as the initial state [1,2,3,4,5,6,7,8,9,10]. On the other hand, if the extreme limit, were considered, where a polypeptide chain departs from its denatured state and in very relatively short period of time finds its unique native state after searching amongst the astronomical number of possible configurations, the simulating process for proteins with fifty to five hundred amino acids using approaches such as Monte Carlo and molecular dynamics, becomes impracticable, due to the very high computation cost. Such contradictory dynamical picture is known as Levinthal paradoxTo investigate the role of stochasticity on the final native state, an inverse strategy is proposed, based on a simple angular 3D random-walk model to build up protein backbones with different lengths and distinct percentages of secondary structures. In the proposed model, each step has a fixed radial size l 0 but dihedral Φ and Ψ angles of the protein backbone are chosen according to independent Gaussian probability distributions, following the suggestion given in reference [12]. The mean value and standard deviation of each d...