When evaluating flutter instability, it is often assumed that incident wind is normal to the longitudinal axis of a bridge and the flutter critical wind speed estimated from this direction is most unfavorable. However, the results obtained in this study via oblique sectional model tests of four typical types of bridge decks show that the lowest flutter critical wind speeds often occur in the yaw wind cases. The four types of bridge decks tested include a flat single-box deck, a flat ${Pi}$-shaped thin-wall deck, a flat twin side-girder deck, and a truss-stiffened deck with and without a narrow central gap. The yaw wind effect could reduce the critical wind speed by about 6%, 2%, 8%, 7%, respectively, for the above four types of decks within a wind inclination angle range between $-3^{circ}$ and $3^{circ}$, and the yaw wind angles corresponding to the minimal critical wind speeds are between $4^{circ}$ and $15^{circ}$. It was also found that the flutter critical wind speed varies in an undulate manner with the increase of yaw angle, and the variation pattern is largely dependent on both deck shape and wind inclination angle. Therefore, the cosine rule based on the mean wind decomposition is generally inapplicable to the estimation of flutter critical wind speed of long-span bridges under skew winds. The unfavorable effect of yaw wind on the flutter instability of long-span bridges should be taken into consideration seriously in the future practice, especially for supper-long span bridges in strong wind regions.