The primary purpose of this study was to investigate the accuracy of the DF50 (ImpediMed Ltd, Eight Mile Plains, Queensland, Australia) bioelectrical impedance analysis device using dual-energy x-ray absorptiometry as the criterion in two groups: endurance athletes and power athletes. The secondary purpose was to develop accurate body fat percentage prediction equations for each group based on bioelectrical impedance analysis data and/or the combination of bioelectrical impedance analysis and anthropometric data. Eighty male athletes (40 elite endurance athletes and 40 were power athletes), age 19–48 with body mass indexes ranging from 18.9 to 37.4 were recruited. Anthropometric measurements were taken. Body composition was assessed by dual-energy x-ray absorptiometry and bioelectrical impedance analysis. An athlete-specific bioelectrical impedance analysis prediction equation was developed by stepwise regression analysis using dual-energy x-ray absorptiometry as the criterion and bioelectrical impedance analysis data and anthropometric measurements as predictor variables. The DF50 bioelectrical impedance analysis significantly overestimated body fat percentage by 6.4 ± 0.5 in the entire group (p < .001) and in both the endurance group (6.1 ± .6, p < .001) and the power group (6.7 ± 0.7, p < .001). The endurance and power group showed no significant difference in the error of estimation by bioelectrical impedance analysis (p = .554), indicating that bioelectrical impedance analysis has the same error in both groups. The final prediction equation incorporated both anthropometric variables as well as bioelectrical impedance analysis variables and produced an adjusted r2 of .982 and a standard error of the estimate (SEE) of 1.98 for the entire group. This prediction equation used bioelectrical impedance analysis measurements and anthropometric measurements, specifically trunk measurements, to account for trunk size, a common source of error in bioelectrical impedance analysis equations. Follow-up validation studies are necessary to further validate the equations produced.