As is well known, the difference between zed and us vector and/or axial vec
tor correlators, whose spectral functions are measurable in hadronic tau de
cay, allows one in principle, through the use of QCD sum rules, to determin
e the strange quark mass, ms. We show that, by studying the behavior of the
relevant correlator difference in the complex q(2)-plane, the freedom to c
hoose arbitrary analytic weight functions present in the finite energy sum
rule (FESR) implementation of this approach may be exploited to construct F
ESR's which (1) significantly reduce theoretical uncertainties and (2) remo
ve problems associated with both the poor convergence of the OPE representa
tion of the longitudinal part of the us vector and axial vector correlators
and the large statistical errors in the us spectral data above the K* regi
on. The result of this optimization is an extraction, based on present data
, m(s) (2 GeV) = 115.1 +/- 13.6 +/- 11.8 +/- 9.7, where the first error is
statistical, the second due to that on V-us, and the third theoretical. We
show also that, for the new weights constructed here, analyses with errors
in the us data in the K* region reduced by a factor of 2 would produce a de
termination of m, with the statistical error reduced significantly below th
at associated with the uncertainty in V-us.