A new strategy of fitting the coupling constants of the nuclear
energy density functional is proposed, which shifts attention
from ground-state bulk to single-particle properties. The latter
are analyzed in terms of the bare single-particle energies and
mass, shape, and spin core-polarization effects.
Fit of the isoscalar spin-orbit and both
isoscalar and isovector tensor coupling constants directly to
the
spin-orbit splittings in
Ca,
Ni,
and
Ca is proposed as a practical realization of this new
programme. It is shown that this fit requires drastic
changes in the isoscalar spin-orbit strength and the tensor coupling
constants as compared to the commonly accepted values but it
considerably and systematically improves basic single-particle
properties including
spin-orbit splittings and magic-gap energies.
Impact of these changes on nuclear binding energies is also
discussed.