When we talk about energy bands, we are talking about mixing of orbitals on different atoms (AOs) with similar energies. First, assume simple models. In the solid, the structure is very regular and all atoms have AOs with identical matching energies. This makes the number of orbitals available for mixing in solids very large, and mixing results in bands. In a liquid, on the other hand, structure is highly irregular. Orbitals on different atoms often have energy or overlap mismatches, but occassionally match and mix, and this results in much local mixing (say due to hydrogen bonding) but not extended delocalized orbitals. Because of energy mismatch you can also end up with bands in liquids but their origin (in this model) is different than in solids.
The difference between the origin of bands in liquids and solids is similar to the different cause of homogeneous and inhomogeneous broadening. In both cases you get bands (a distribution of energies), but in one case broadening involves delocalization over resonantly matched orbitals on different atoms/molecules and in the latter it involves variations in the local environment due to randomly fluctuating perturbations caused by neighboring molecules ("the solvent"). Extensive orbital mixing and Pauli's exclusion principle results in homogeneous broadening in the first case. Motional jostling in the case of a liquid results in inhomogeneous broadening.
Energy levels of individual molecules in liquid water are "perturbed" by neighboring molecules. These perturbations are small compared to the ionization energy of the valence electrons. The stability conferred by an H-bond in water is in the order of 20 kJ/mol. Compare that to the ionization energy, ~1200 kJ/mol. Therefore electrons remain bound to local waters even as their energies are slightly perturbed by interactions with other molecules. H-bonds do contain covalent character, that is, involve sharing of electrons between two molecules (specifically between H-bond donor and acceptor atoms). However this does not imply electrons in liquid water are extensively delocalized beyond local bonds, certainly not on the spatial scale possible in solids.
As a note of interest ice under pressure exhibits metallic phases.
Note perturbations due to the solvent are evident in the difference between vibrational spectra obtained in different phases.