When one plots the locations of some not too biased sample of stars in a diagram of absolute magnitude and spectral type (a so-called Hertzsprung-Russell diagram), one finds that most of the stars lie on the main sequence, a gently bent line from upper left to lower right.
This line is so pronounced because both the spectral type and the luminosity depend on a star's mass only to zeroth order as long as it is fusing hydrogen -- and that is what almost all stars spend most of their "active" life doing.
At closer inspection, one notices that the main sequence is not exactly a line but instead somewhat fuzzy. There are many reasons for this fuzziness, the most important one still being observational uncertainties which mainly affect the distance of the star in question but range all the way to unresolved binary stars.
But even perfect observations would lead to a fuzzy main sequence, because mass is not the only parameter a star has after all. Chemical compositon and -- related -- its evolutionary status also move a star slightly on the main sequence, as do close companions, rotation, or magnetic fields, to name just a few. Actually, there are very metal-poor stars (subdwarfs) that to lie just below the main sequence although they are fusing hydrogen, thus marking the lower edge of the main sequence's fuzzyness due to chemical compositon.
Stars usually enter and leave the main sequence from above when they are born or they are starting to die, respectively.