A \alert{one-step} strategy maps every reducible term to a set of one-step reductions.
  There exists no fair one-step strategy for $\atrs = \{I(x) \to I(x)\}$.

  For the term:  
    t = I(I(x))
  there are only 3 possible mappings:
    \item $\SS(t) = \{\alert{I}(I(x)) \to_{\varepsilon} I(I(x))\}$,
    \item $\SS(t) = \{I(\alert{I}(x)) \to_{1} I(I(x))\}$, or
    \item $\SS(t) = \{\alert{I}(I(x)) \to_{\varepsilon} I(I(x)), \; I(\alert{I}(x)) \to_1 I(I(x))\}$.
  None of these is fair as we can always continue to reduce the same occurrence of $I$.