Over 12 years of IMP 8, data was searched for observed bow shock crossings. Out of the total 4562 crossings found, we used the 2293 unambiguous bow shocks for which upstream interplanetary magnetic field and solar wind parameters were available to study selected bow shock models under normal and unusual solar wind conditions. The chosen models were F79, NS91, FR94, FR94c, CL95, and P95 <Formisano, 1979; Nemecek and Šafr¿nkov¿, 1991; Farris and Russell, 1994; Cairns and Lyon, 1995; Peredo et al., 1995>. This statistical study investigates these models' reliability not only for average solar wind plasma and interplanetary magnetic field (IMF) conditions but also for unusual conditions and as a result points out some deficiencies of these models. Statistically, the predictions of F79 and the phenomenological and MHD models FR94, FR94c, and CL95 are the most accurate, with F79 giving a slightly better result. The P95 model predicts standoff distances which are too large by ~20%. For large values of the IMF and its components, all models except NS91 underestimate the bow shock distance. Furthermore, the models underestimate the bow shock distance when the upstream Mach numbers are low (≲5). The models also do not properly reflect changes in the relative orientation of the IMF and solar wind velocity vectors. An independent evaluation of the dawn and dusk sectors suggests an asymmetry in the bow shock shape and/or a different reaction of the flanks to solar wind deviations from a radial flow. Taking the upstream parameters from a distant solar wind monitor (the Wind spacecraft) resulted in the models predicting the shock farther away from the Earth, which is likely a result of the spacecraft separation perpendicular to the solar wind flow, or of calibrational differences of the plasma density measurements by the spacecraft.