In any theory it is unnatural if the observed parameters lie very close to special values that determine the existence of complex structures necessary for observers. A naturalness probability, P, is introduced to numerically evaluate the unnaturalness. If P is small in all known theories, there is an observer naturalness problem. In addition to the well-known case of the cosmological constant, we argue that nuclear stability and electroweak symmetry breaking (EWSB) represent significant observer naturalness problems. The naturalness probability associated with nuclear stability is conservatively estimated as P_nuc < 10^{-(3-2)}, and for simple EWSB theories P_EWSB < 10^{-(2-1)}. This pattern of unnaturalness in three different arenas, cosmology, nuclear physics, and EWSB, provides evidence for the multiverse. In the nuclear case the problem is largely solved even with a flat multiverse distribution, and with nontrivial distributions it is possible to understand both the proximity to neutron stability and the values of m_e and m_d - m_u in terms of the electromagnetic contribution to the proton mass. It is reasonable that multiverse distributions are strong functions of Lagrangian parameters due to their dependence on various factors. In any EWSB theory, strongly varying distributions typically lead to a little or large hierarchy, and in certain multiverses the size of the little hierarchy is enhanced by a loop factor. Since the correct theory of EWSB is unknown, our estimate for P_EWSB is theoretical. The LHC will determine P_EWSB more robustly, which may remove or strengthen the observer naturalness problem of EWSB. For each of the three arenas, the discovery of a natural theory would eliminate the evidence for the multiverse; but in the absence of such a theory, the multiverse provides a provisional understanding of the data.

79 pages, 23 figures