The aims of this study were (a) to determine which of the most used anthropometric equations was the most accurate to estimate percentage of body fat (%BF), (b) to develop a new specific anthropometric equation, and (c) to validate this football-specific equation. A total of 126 (13.3 ± 0.6 years) football players (86 males and 40 females) participated in the present study. Participants were divided into two groups: 98 players were included in the assessment of existing equations and in the development of the new prediction equation, and 28 players were used to validate it. %BF was measured with dual-energy X-ray absorptiometry (DXA) and also estimated with six different %BF anthropometric equations: Johnston, Slaughter, Carter, Faulkner, Deurenberg, and Santi-Maria. Paired t tests were used to analyze differences between methods. A football-specific equation was developed by a stepwise linear regression. The existing anthropometric equations showed significant bias for %BF when compared with DXA (p < .001; constant error ranged from −4.57% to 9.24%; standard error of estimate ranged from 2.46 to 4.20). On the other hand, the developed football-specific equation was %BF = 11.115 + 0.775 (triceps skinfold) + 0.193 (iliac crest skinfold) − 1.606 (sex). The developed equation demonstrated neither %BF differences (p = .121; constant error = 0.57%; standard error of estimate = 0.36) when compared with DXA, presenting a high cross-validation prediction power (R 2 = .85). Published anthropometric equations were not accurate to estimate %BF in adolescent football players. Due to the fact that the developed football-specific equation showed neither differences nor heteroscedasticity when compared with DXA, this equation is recommended to assess %BF in adolescent football players.