Assessing the efficacy of anticancer agents in animal models remains a necessary step in the development of new treatment options and plays an important role in their optimization and comparison. Often, however, interpretation of the results is flawed by excessive trust in scores traditionally handed down, but whose origin and limitations have been lost. Here I examine the theories and assumptions underlying the most common rating scales, suggesting improvements to the old scores and proposing the adoption of multi-parameter analysis and interpretation of the results, considering different time-windows. I examined case examples of different scenarios of antiproliferative effects induced by treatment, demonstrating that common scores fail to distinguish between completely different responses to treatment or, in other circumstances, indicate a different outcome when the response is the same. I found that a combination of parameters, including the percent tumor growth between the start and end of treatment, the relative tumor burden at nadir and the absolute growth delay, may distinguish among the different cases and support a correct interpretation of the antitumor response. All these parameters can be derived from individual tumor growth curves in a simple way, without any change to common experimental procedures.