Search is an important aspect of Artificial Intelligence and many advances have been achieved in finding optimal solutions for a variety of search problems. Up until recently most search problems were solved using a serial-single threaded approach. Speed is extremely important and one way to decrease the amount of time needed to find a solution is to use better hardware. A single threaded approach is limited in this way because newer processors are not much faster than previous generations. Instead industry has added more cores to allow more threads to work at the same time. In order to solve this limitation and take advantage of newer multi-core processors, many parallel approaches have been developed. The best approach to parallel search is an algorithm named Parallel Best-N Block First Search (PBNF). PBNF relies on an abstraction function to divide up the work in a way that allows threads to work efficiently with little contention. This thesis studies the way this abstraction function chooses to build the abstraction and demonstrates that better abstractions can be built. This abstraction focuses on goal variables on ways to keep the number of abstract states as small as possible while adding as many variables as feasible.