We develop empirical predictive equations for shear‐wave velocity (VS) conditional on standard penetration test resistance (N‐value) and other explanatory variables using data sets from Japan. The VS and N data sets are from the Kyoshin network (K‐NET), which include 16,845 collocated measurements of VS and N at 1102 sites. We begin with baseline equations considering soil type and overburden pressure in addition to N. For coarse‐grained soils, VS is more sensitive to overburden pressure than to N, whereas for fine‐grained soils, VS is more sensitive to N. We find that residuals (difference between data and model) for a given borehole tend to be correlated (e.g., samples within a boring are consistently high or low); hence, we use mixed‐effects regression to compute a quantity akin to an average borehole residual referred to as the between‐boring residual. Sites associated with older (pre‐Holocene, >10,000 yrs) geology tend to have null to slightly positive between‐boring residuals, whereas these residuals are negative for more recent materials. Accordingly, we provide adjustments to the baseline equations conditional on geologic condition. Between‐boring residuals exhibit spatial correlations; however, due to lack of knowledge as to the cause of these correlations, we do not recommend such effects for inclusion in the model. When applied to predict the time‐averaged VS in the upper 30 m (VS30) in the absence of direct measurement, the proposed equations provide significantly better accuracy than widely used local geomorphology‐based proxies (0.26 versus 0.42 of standard deviation of the natural log of residuals). This suggests that penetration resistance data can improve predictions of VS30 compared with geomorphology‐based proxies alone, when site‐specific VS measurements are unavailable.