Using the predictions of the Lynden Bell/Hills model of the energy source of QSO's in which discrete packets of matter (e.g., stars) are converted into radiation near a massive black hole, it is shown how the variability observations of QSO's over long periods may be used to upgrade the Hubble diagram for these objects. In a class of objects of low shot-rate, extensive simulations using a shot noise model indicate that individual events may be distinguishable. This enables the calibration of the relation between shot-rate and the variance of apparent luminosity, and hence the derivation of shot-rates for objects with measured light curves. From these shot-rates, absolute luminosities are in principle derivable, and here the observational difficulties still impeding progress in path towards reliable values of g(O) are indicated using this method.