In order to highlight the sources of economically valuable traits in 2017-2018 years, 131 varieties of pea collection (Pisum sativum L.) were clustered using a hierarchical agglomerative algoritm based on the minimum of Euclidean distanced. Under conditions with sufficient moisture (2017, GTC=1.35) and moisture deficit (2018, GTC=0.76) in two morphological groups different in leaf type, the varieties were combined into clasters according to similar yield values (g/m2), the protein content in seeds (%) and the duration of the growing season (day) from full germinition to economic maturation. In 2017, in groups with a leafless and a common type leaf, eight clusters were distinguished in each of them, in arid conditions (2018), their number was, respectively, nine and seven. For use in breeding to increase the gross yield of protein, clusters with high yields and protein content are proposed, mainly in the leafless group, which have a higher resistance to lodging. In this morphological group, under conditions with suficient moisture varieties of the sixth cluster Yamalsky, Yamal-2, Aksaysky 55, Stepnyak, Krasnoufimskiy 11, Berkut, Terno with a yield of 380-492 g/m2 and a protein content of 23.23...27.81%, in arid conditions (2018) the varieties of the fifth cluster Faraon, Fokor, L-1599, Pamjat Khangildina, Samarius with indicators, respectively, 268...312 g/m2) and 22.61...22.93% were distinguished. In 2017 among the leaf varieties of the fourth cluster Kudesnik, Argon, L-2516 with the combination of high values of productivity (400...428 g/m2) and protein content in seeds (23.67...25.32%), UG-95888-2, in 2018 varieties of the fifth cluster Intensive 92, Janus, Chishminsky 229 with indicators of 268...320 g/m2 and 22.51...25.42% were distinguished. Varieties Veles and Kazanets with consistently high yields (304-328 g/m2, CV = 0.6 ... 1.3%) are of breeding value as sources of drought resistance