<p>The usual asset pricing models perform well when large datasets are available, particularly with the advent of big data and artificial intelligence. However, they are generally unable to meet current assessment needs when only few data are available. For example, in agricultural land valuation, farmland transactions are generally non-recurring actions for most buyers and sellers, and expert knowledge is often the sole source of information, posing significant challenges for decision-makers. A valuation method based on two distribution (or reliability) functions was developed to estimate asset value using only expert-based knowledge of the asset and its characteristics. However, this comparative method and its multidimensional extensions tend to produce extreme valuations. This paper introduces competing risks and complementary risks models within this probabilistic framework to support asset valuation under high uncertainty and data scarcity. The proposed method synchronizes the asset value distribution with the minimum and maximum statistics of the explanatory variables, yielding more moderate and realistic valuations. The results confirm that embedding these risk models refines asset assessments by mitigating the appraisal biases induced by the two-function valuation methods across any <i>k</i>-dimensional explanatory vector. Empirical evidence from two applications in agricultural land valuation supports the theoretical findings. The results show minimal impact of the shape parameters on valuation outcomes and limited sensitivity to the expert’s most likely value, exhibiting reasonable performance. Accordingly, the minimum and maximum statistics can serve as stochastic smoothing tools for valuation, thereby extending the applicability of valuation methods in data-scarce markets.</p>