The most promising mechanism for the formation of contact binaries involves the orbital angular momentum loss (AML) and the resulting orbital decay of detached but close synchronized binaries. The efficiency of magnetic wind braking should abruptly decrease upon formation of a contact binary because of the transformation into a system of earlier spectral type and (possibly) of longer orbital period. The new primary of the contact system should have convective zone thinner than indicated by the surface temperature of the common envelope. The decrease in the coronal (X-rays and radio) activity of contact binaries, which is indeed observed, is used as an agrument that the AML efficiency in contact is relatively low and that the contact stage is considerably prolonged relative to adjacent stages. This small modification to the AML models is capable of explaining why many different contact binaries are observed in old systems like NGC188. The AML evolution is not the only mechanism leading to formation of contact binaries; some of them may have originated via Algol-like evolution. Thus, the observed contact binaries are probably a mixture of systems formed in different ways.