Traffic signal timings can be optimized with offline or online tools. Offline tools, such as Synchro, take a macroscopic approach, whereas genetic algorithm (GA) formulations rely on a close interaction with traffic microsimulators. Online tools, such as the Split-Cycle Offset Optimization Technique (SCOOT), optimize in real time. Offline optimization tools have the luxury of time for repetitive computation, whereas SCOOT must work quickly as it responds to traffic detected on the streets. A comparison of the best offline tools and adaptive signal control is presented. The signal timing plans are first derived from the Synchro macroscopic optimization tool. A genetic algorithm (GA)-based formulation, which is essentially stochastic, resides in the VISSIM traffic simulation software; the GA component is known as VISGAOST. The latest version of SCOOT (MC3) is modeled with upstream and stop line detectors to enable both phase skipping and overlaps. The test bed is closely modeled on a 14-intersection network in Park City, Utah. The quality of signal timings is evaluated for two traffic demand types: expected (no significant variation) and unexpected (random variations). Results indicate that the offline GA formulation provides the optimal signal timing plans, robust enough to accommodate even randomly changing traffic demand. Similar in performance to Synchro, SCOOT delivers a creditable quality of optimization for an online optimization tool.