Of course I never write my own Hash Table implementation, nor should I. But this is fascinating anyway -- and obvious in retrospect. "Reduce the length of the longest probe chains by reorganizing the probe chains to even them out. This improves performance when the table is full enough that the probe chain length matters."