Modern web applications are usually based on JavaScript. Due to its loosely typed, dynamic nature, test execution is time expensive and costly. Techniques for regression testing and fault-localization as well as frameworks like the Google Web Toolkit (GWT) ease the develop- ment and testing process, but still require approaches to reduce the testing effort. In this paper, we investigate the efficiency of a spe- cialized, graph-walk based selective regression testing technique that aims to detect code changes on the client side in order to determine a reduced set of web tests. To do this, we analyze web applications created with GWT on different precision levels and with varying looka- heads. We examine how these parameters affect the localization of client-side code changes, run time, memory consumption and the num- ber of web tests selected for re-execution. In addition, we propose a dynamic heuristics which targets an analysis that is as exact as possible while reducing memory consumption. The results are partially appli- cable on non-GWT applications. In the context of web applications,we see that the efficiency relies to a great degree on both the structure of the application and the code modifications, which is why we propose further measures tailored to the results of our approach.