ABSTRACTWheat yield is influenced by fertilisation, precipitation and variety, among other factors. There is limited research identifying the most important factors affecting wheat yield and assessing their relative importance in the long run. In this study, we evaluated the contribution of fertilisation, precipitation and variety to wheat yield using a long-term field experiment (1984–2014) on the semiarid Loess Plateau in China. The experiment consisted of six treatments: fertilisation with nitrogen (N), phosphorus (P), manure (M), NP, NPM, and a control without fertilisation. We monitored the yield of three varieties of winter wheat over time and assessed the changes in grain yield, soil properties, fertiliser-contribution rate (FCR) and precipitation-use efficiency (PUE) with different fertilisation treatments and precipitation patterns. Stepwise multiple linear regression was used to identify the most important factors affecting wheat yield and examine their relative importance. The results showed that fallow-season precipitation and annual precipitation (AnP) positively correlated with wheat yield in the N, M, NP, and NPM treatments. The amount of fertilisation, AnP, and monthly precipitation of February and September were included in the linear regression model; however, the influence of variety on yield could be ignored. With 30 years of fertilisation, soil organic matter, total nitrogen, and available potassium levels with NPM was higher than the control by 70.6%, 70.5%, and 319.2%, respectively. Yield, FCR, and PUE with M increased annually at rates of 89 kg ha−1 y−1, 1.47 kg kg−1 y−1, and 0.13 kg mm−1 y−1, respectively. The yields and FCR, but not PUE, of all fertilised treatments were higher in wet than normal and dry years. The FCR with P was negative in all the three precipitation patterns. This study has implications for maximising the long-term winter yield with various factors in the rain-fed winter wheat cropping system of the Loess Plateau.