Bioelectrical impedance analysis (BIA) is a common practice to assess body composition in athletes, however, when measuring athletes with specific body geometry, its accuracy may decrease. In this study we examined how length dimensions affect body composition estimation and we compared BIA and dual-energy X-ray absorptiometry (DXA) assessments in three sports. 738 male adolescent athletes (15.8 ± 1.4 years) from three sports (soccer, basketball, and handball) were measured. Body composition was estimated by BIA (InBody 720) and by DXA (Lunar Prodigy). Differences between the two methods were tested by Bland-Altman analysis and by paired t-test. ANOVA was used for inter-group comparisons. Pearson correlation and multivariate linear regression was used to look for the relationship between segmental lean body mass and length dimensions. BIA_{InBody 720} consistently underestimated percent body fat (PBF) and overestimated lean body mass (LBM) than DXA. The magnitude of the differences between the two methods varied among the examined sports. Handball (PBF = 8.3 ± 2.4 %; LBM = -5.0 ± 2.1 kg) and basketball players (PBF = 8.8 ± 2.3 %; LBM = -5.3 ± 1.8 kg) had significantly larger differences between the two methods than soccer players (PBF = 6.4 ± 2.2 %; LBM = -3.1 ± 1.4 kg). There was a negative correlation between differences in segmental LBM estimation and length sizes (trunk length, upper extremity length, lower extremity length). The highest correlation was found for lower extremity (r = -0.4). Longer lower extremity resulted in greater difference in LBM estimation. The differences between the sport disciplines are most probably attributed to body height differences. Length dimensions result in overestimation of LBM with BIA, thus body composition assessment with BIA_{InBody 720} needs to be carefully interpreted in athletes with extreme length sizes, especially, with basketball players.