As software systems become increasingly complex, ensuring correct and efficient run-time behavior is a significant challenge. Runtime monitoring addresses this challenge by verifying the behavior of the program during execution by checking the specifications against the current state of the program. RTLola, a stream-based monitoring language, targets this problem by defining specifications through stream equations on input streams. Output streams compute new values based on both current and past data, while triggers indicate specification violations, allowing an immediate response. However, we also want to compute values based on future data, so that we can handle mixed temporal properties. This allows us to predict the behavior of the system and respond accordingly, closing the gap between what has happened and what is expected. The idea of future offsets first came from Lola, a synchronous monitoring language. The move from Lola to the asynchronous approach of RTLola was necessary to support modern systems that require real-time and flexible monitoring. RTLola focuses on efficiently handling current and past data, which is why future offsets are not currently handled. This thesis proposes an extension to RTLola by integrating the future offset operator, combining the asynchronous and real-time data access features of RTLola with the concepts of future offsets originally introduced in Lola. To demonstrate the applicability of this extension, we revisit the fundamental principles of Lola and present theoretical analysis alongside a practical implementation within the RTLola framework. The evaluation ensures that each set of input streams has a unique evaluation model and that the design prevents computational inefficiencies by keeping the worst-case memory usage constant relative to the trace size.