The load/unload response ratio (LURR) is a short-to-intermediateterm earthquake prediction method. Using this method, the LURR time series usually climb to an anomalously high peak prior to occurrence of a large earthquake. Previous studies have indicated that the size of critical seismogenic region selected for LURR measurements has great influence on the evaluation of LURR. We replace the circular region usually adopted in LURR practice with an area, within which the tectonic stress change would mostly affect the Coulomb stress on a potential seismogenic fault of a future event. This new algorithm is devised to improve the predictive power of the LURR method. The Coulomb stress change before a hypothetical earthquake is calculated based on a simple backslip dislocation model of the event. Retrospective tests of this algorithm on four M＞6.5 earthquakes occurred in California over the last two decades and the 2008 MS8.0 Wenchuan earthquake, show that the anomalies produced by the Coulomb stress algorithm look more prominent than the ones produced by the circular region algorithm. Moreover, The Coulomb stress algorithm seems to be more sensitive for detecting future earthquakes of particular magnitude range and location than the circular region algorithm. The unique feature of the Coulomb stress algorithm may provide information and constraints on forecasts of the size and location of future large events based on the regional tectonic activities.