【Objective】 The best linear unbiased prediction (BLUP) method forecasts a breeding value using a mixed linear model equation, which reflects true genetic effects with high accuracy. 【Method】 This study used two open pollination progeny tests of Picea crassifolia Kom. at 13 and 14 years old obtained from a primary clonal seed orchardto evaluate genetic variation in height. A breeding value was predicted using the BLUP method. 【Result】Our results showed that family and interaction effects between family and blocks significantly affected height. The family hereditability for height of the trees from the two open pollination progeny were 0.17 and 0.36, respectively, while the individual hereditability were both 0.07. The Pearson correlation between the original value and the breeding value of the height at family level, or at individual level, were both high, and the Pearson correlation coefficients were 0.99 and 0.86, respectively. According to the breeding value of the parents of the families, 20 elite families were chosen and the reality genetic gain reached 36% and 26%, respectively. According to the individual breeding value estimated using the BLUP method, 20 plus trees were chosen, and the expected genetic gains were 7.6% and 7.9%, respectively. 【Conclusion】By comparing the genetic gain of backward selection and forward selection, it was determined that backward selection is the better choice in high-generation breeding ofPicea crassifolia.