The hash size (code) is also known as the k-tuple size or the word size. When the analysis initially scans the sequences for matches, it uses a very fast technique known as hashing. When the hash size is set to six, the program will only consider exact matches of six residues to be a hit. If it is set to one, it will consider a single residue match to be a hit, and so on. Using larger hash sizes will result in a less sensitive but faster search.
In other words, the hash value is a measure of how long an exact match between two sequences must be before the dot matrix algorithm will attempt to score and align that matching region. A hash value of 1 is the most sensitive, 6 is the least sensitive.
TIP: For most comparisons in applications like MacVector, start with a hash value of 6, because it is unusual for two sequences to possess significant similarity without having regions of that size that match exactly.
Reference:
MacVector 7 User Guide. Oxford Molecular Ltd., 2000.