Aim To compare the effects of 16‐week multicomponent and resistance training, and 6‐week detraining on physical variables related to a higher risk of falls in very old people. Methods A randomized, three‐arm, controlled trial was carried out. A total of 69 community‐dwelling older adults aged 80 years and older were allocated to three groups: control, multicomponent training and resistance training. They were assessed at baseline, after 16‐week training and 6‐week detraining. The control group did not perform any intervention. The multicomponent group performed protocol consisting of warm‐up, aerobic, strength, balance and cool‐down exercises. The resistance group underwent strength exercises using six adapted machines. The training sessions had progressive intensity, lasted 16 weeks and 12 included three 1‐h sessions per week. The assessment consisted of anamneses, five‐repetition sit‐to‐stand, one‐leg standing, tandem and dual task tests. For statistical analysis, α = 0.05 was used. Results There were no significant differences between groups and assessments in any variable when analyzed by intention to treat. However, when analyzed, the older adults who adhered to the training, the multicomponent group, had a significant improvement in the sit‐to‐stand and the one‐leg standing (right support) tests. There was a significant main effect between times on the one‐leg standing (left support) test. Conclusion In very old people, multicomponent training seems to be more beneficial and presents fewer adverse events when the adherence to protocol is higher. Geriatr Gerontol Int 2016; 16: 492‐499.