Observation Based Slicing is a program slicing technique that depends purely on the observation of dynamic program behaviours. It iteratively applies a deletion operator to the source code, and accepts the deletion (i.e. slices the program) if the program is observed to behave in the same was as the original with respect to the slicing criterion. While the original observation based slicing only used a single deletion operator based on deletion window, the catalogue of applicable deletion operators grew recently with the addition of deletion operators based on lexical similarity. We apply a hyperheuristic approach to the problem of selecting the best deletion operator to each program line. Empirical evaluation using four slicing criteria from Guava shows that the Hyperheuristic Observation Based Slicing (HOBBES) can significantly improve the effeciency of observation based slicing.