How representative is that of your actual data? It looks from that sample as though you can start at the third character, then find the first number after that, and see if there are at least 3 characters in between, rather than having to look for a sequence of at least 3 amongst many possible sequences.