The experimentally postulated mechanism for the interconversion between (S)-vinylglycolate and (R)-vinylglycolate catalyzed by mandelate racemase enzyme consists of a two-step quite symmetric process through a dianionic enolic intermediate that is formed after the abstraction of the α-proton of vinylglycolate by a basic enzymatic residue and is then reprotonated by another residue. The challenging problem behind this reaction is how the enzyme manages to stabilize such an intermediate, that is, how it lowers enough the high pK a of the α-proton for the reaction to take place. The QM/MM simulations performed in this paper indicate that catalysis is based on the stabilization of the negative charge developed on the substrate along the reaction. We have identified three different reaction mechanisms starting from different quasi-degenerate structures of the substrate - enzyme complex. In two of them the stabilizing role is done by means of a catalytic proton transfer that avoids the formation of a dianionic intermediate, and they involve six steps instead of the two experimentally proposed. On the contrary, the third mechanism passes through a dianionic species stabilized by the concerted approach of a protonated enzymatic residue during the proton abstraction. The potential energy barriers theoretically found along these mechanisms are qualitatively in good agreement with the experimental free energy barriers determined for racemization of vinylglycolate and mandelate. The theoretical study of the effect of the mutation of Glu317 by Gln317 in the kinetics of the reaction reveals the important role in the catalysis of the hydrogen bond formed by Glu317 in the native enzyme, as only one of the mechanisms, the slower one, is able to produce the racemization in the active site of the mutant. However, we have found that this hydrogen bond is not an LBHB within our model.