We compare 237 Lyman-α (Lyα) spectra of the "MUSE-Wide survey" ) to a suite of radiative transfer simulations consisting of a central luminous source within a concentric, moving shell of neutral gas, and dust. This six parameter shell-model has been used numerously in previous studies, however, on significantly smaller datasets. We find that the shell-model can reproduce the observed spectral shape very well -better than the also common 'Gaussian-minus-Gaussian' model which we also fitted to the dataset. Specifically, we find that ∼ 94% of the fits possess a goodness-of-fit value of p(χ 2 ) > 0.1. The large number of spectra allows us to robustly characterize the shell-model parameter range, and consequently, the spectral shapes typical for realistic spectra. We find that the vast majority of the Lyα spectral shapes require an outflow and only ∼ 5% are well-fitted through an inflowing shell. In addition, we find ∼ 46% of the spectra to be consistent with a neutral hydrogen column density < 10 17 cm −2 -suggestive of a non-negligible fraction of continuum leakers in the MUSE-Wide sample. Furthermore, we correlate the spectral against the Lyα halo properties against each other but do not find any strong correlation.