A new method is presented to identify transient candidates in large-scale surveys based on the variability pattern in their light curves. The method is based on the Abbe value, A, that estimates the smoothness of a light curve, and on a newly introduced value called the excess Abbe and denoted EA, that estimates the regularity of the light curve variability pattern over the duration of the observations. The new method proves to be a potentially powerful tool to extract transient candidates from large-scale multi-epoch surveys. The better the photometric measurement uncertainties are, the cleaner the list of detected transient candidates is. In addition, the A versus EA diagram is shown to be a good diagnostic tool to check the data quality of multi-epoch photometric surveys. A trend of instrumental and/or data reduction origin, for example, will manifest itself by an unexpected distribution of points in the diagram.