The Physalis angulata L. species, better known in Brazil as camapu, is most found in the northern regions of the country. It is used for human consumption, in medicine and as an ornamental plant. Because of its potential economic and nutritional value, studies have carried out to known the genetic variability of fruit traits. Repeatability is the fraction of phenotypic variation that is due to permanent genetic and environmental factors. The estimation of repeatability coefficient allows breeders to estimate the ideal number of measurements for a trait. This work aims to evaluate the performance of P. angulata accessions and to estimate the coefficients of repeatability and minimum number of repetitions for fruit characteristics, in two environments and two seasons. The present work was conducted at the Horto Florestal Experimental Unity of Universidade Estadual de Feira de Santana, using seven accessions of P. angulata from the Germplasm Collection of LAGEM/UEFS. Four experiments were conducted, two in the experimental field and two in the greenhouse, in an entirely randomized design with 24 replicates, using five accessions in the first two experiments, and six accessions in the last two in each environment, totaling 120 and 144 experimental units, respectively. The characteristics number of fruits per plant (NFP), soluble solids content (SSC), longitudinal diameter (LFD) and transversal diameter (TFD) of the fruits, and average fruit mass (AFM) were evaluated. Data were submitted to analysis of variance (ANOVA) and estimates of repeatability coefficients were obtained by ANOVA, principal components (PC), and structural analysis (SA) methods. There was significant variance among accessions for the characteristics LFD, TFD, AFM and SSC in the experimental field, and for TFD, LFD NFP in the greenhouse. Heritability ranged from 61.63–91.91% in both environments. PC showed the highest estimates of repeatability coefficient. The repeatability and determination coefficients by PC analysis ranged from TFD (r = 0.51; R2 = 96.19%) to SSC (r = 0.72; R2 = 98.47) in the field, and from NFP (r = 0.40; R2 = 93.92%) to LFD (r = 0.51; R2 = 79.38%) in the greenhouse. Twenty measurements can achieve 95% and 90% reliability for experimental field and greenhouse experiments, respectively, for the evaluated traits.