degenerate oligo design

[Tom MacAllister]

Dear Netters,
      I wish to design a degernate oligo based on peptide sequence for
screening a genomic library.  I wondered if anyone could give me an idea as
to the relative importance of number of possible combinations vs number of
known bases.  I have the follosing options:

          20mer     2048 combinations     14 known bases
          17mer      512 combinations     12 known bases
          14mer      128 combinations      9 known bases

Is it better to go with more known bases and hope that a lot of the degenerate
combinations will hybridize anyway?  My feeling is that the longer oligo would
give me more room to mess with hybridization conditions.
                                           Tom