The Collaboratory for the Study of Earthquake Predictability (CSEP) is a global project aimed at testing earthquake forecast models in a fair environment. Various metrics are currently used to evaluate the submitted forecasts. However, the CSEP still lacks easily understandable metrics with which to rank the universal performance of the forecast models. In this research, we modify a well-known and respected metric from another statistical field, bioinformatics, to make it suitable for evaluating earthquake forecasts, such as those submitted to the CSEP initiative. The metric, originally called a <em>gene-set enrichment score</em>, is based on a Kolmogorov-Smirnov statistic. Our modified metric assesses if, over a certain time period, the forecast values at locations where earthquakes have occurred are significantly increased compared to the values for all locations where earthquakes did not occur. Permutation testing allows for a significance value to be placed upon the score. Unlike the metrics currently employed by the CSEP, the score places no assumption on the distribution of earthquake occurrence nor requires an arbitrary reference forecast. In this research, we apply the modified metric to simulated data and real forecast data to show it is a powerful and robust technique, capable of ranking competing earthquake forecasts.