Oxide-based hydrogenation catalysts have attracted intensive interest, but the relationships between their composition, structure, and reaction mechanism are still ambiguous. Here, we conducted density functional theory (DFT) calculation on ethylene hydrogenation over WO 3 , WO 2.72 , WO 2 , and W catalysts to explore how the structure and catalytic activity change with the composition and explain why neither WO 3 nor metallic W is a good hydrogenation catalyst but WO 3−x is. Calculations on the geometric and electronic structures show a transition from semiconductor to metal-like with more W−W metal bonds appearing from the corresponding oxidized to metallic tungsten. Correspondingly, the H 2 dissociation mechanism changes from heterotypic on the semiconductor WO 3 and WO 2.72 to homotypic on the metal-like WO 2 and metallic W, and the adsorption strength of the dissociation product (2H) enhances from oxidized to metallic W. Calculations on the stepwise hydrogenation indicate that H 2 dissociation is the rate-limiting step (RLS) for WO 3 and WO 2.72 , whereas the following hydrogenation step (C 2 H 5 + H → C 2 H 6 ) is the RLS for WO 2 and W. By linear fitting the binding energies of H 2 , 2H, H + C 2 H 5 , and C 2 H 6 adsorptions, a perfect correlation between (H + C 2 H 5 ) and (2H) adsorptions is observed; thus, the catalytic activity and mechanism can be evaluated using 2H adsorption behavior as the descriptor. On the basis of the calculation results, we predict that the best ethylene hydrogenation activity will show at 2 < x < 2.72 for WO x . This work suggests that controlling the composition between the stoichiometric oxide and the metal may be an effective way to develop an active hydrogenation catalyst.