FFMI calculator

How it works

Fat-free mass is everything that is not fat: muscle, bone, organs, water. FFMI divides it by height squared, as BMI does with weight.

• fat-free mass = weight × (1 − body fat % / 100)
• FFMI = fat-free mass (kg) / height (m)²
• normalised FFMI = FFMI + 6.3 × (1.80 − height in m)

Taller lifters score slightly higher on plain FFMI. The normalised form corrects for that by adjusting every result to a height of 1.80 m: plus 0.63 at 1.70 m, minus 0.63 at 1.90 m.

Kouri and colleagues introduced the correction in 1995, and their abstract gives 6.3. Some calculators use 6.1. The gap is 0.02 for every 10 cm from 1.80 m: 0.04 at 1.60 m or 2.00 m. Kouri studied men only. We apply the correction to women for comparison only.

The range

We run the formulas again at your body fat plus and minus the typical error of your method. More fat, less fat-free mass. So the low end comes from the higher body fat. The error is one standard deviation, which means about two in three measurements land inside it.

Why the body-fat error dominates

Height and weight are measured to a centimetre and a few hundred grams. Body fat is not.

At 1.80 m and 85 kg, each body-fat point moves FFMI by 0.26 (0.85 / 3.24). A DXA scan's ±3 points is ±0.8 FFMI. A bathroom scale's ±5 is ±1.3. Two identical lifters on different scales can land either side of a reference point.

Error by method, from validation studies:

• DXA: read 1.8 points below a four-compartment model on average. Individual gaps ran from −2.6 to +7.3, and lean people read lowest (Van Der Ploeg 2003, 152 adults).
• BodPod and underwater weighing: total error 2.7 and 2.5 points against a three-compartment model, 31 college men (Moon 2008).
• Lab or 8-electrode BIA: 3.3 points total error for the two 8-electrode devices among 15 tested (Siedler 2023).
• Skinfold calipers: 2.4 points total error against underwater weighing in 35 lean men (Eckerson 1992). For a change in body fat in elite judo athletes, limits of agreement ran −3.4 to +3.6 (Silva 2009).
• Bathroom scales and handheld BIA: total errors up to 7.8 points, 5.4 for a hand-to-hand device (Siedler 2023). Three smart scales missed DXA by 4.5 to 5.9 on average (Majmudar 2022).
• Photos and visual estimates: the best-tested phone photo method had 95% limits of −5.5 to +4.7 against DXA (Majmudar 2022). Two expert raters judging lean men had 2.3 points total error for one rater, and differed from each other by 2.7 on average (Eckerson 1992). We found no study of people judging their own photos. We use ±5, as for a bathroom scale.

What the reference points were

• Schutz 2002: 5,635 Swiss adults, body fat by BIA. Median FFMI at 18 to 34: 18.9 for men, 15.4 for women. Plain FFMI, general population.
• Kouri 1995: 157 male athletes, 83 steroid users and 74 non-users. The non-users' normalised FFMI "extended up to a well-defined limit of 25.0". That is the top of one sample of 74 men, measured by skinfolds. Many users in the same study went past 25. Twenty Mr America winners from 1939 to 1959 averaged 25.4, on estimated body fat.
• Trexler 2017: 235 college American football players, DXA, normalised to 1.80 m. 26.4% were above 25. The 97.5th percentile was 28.1.

Above 25 means above the highest of Kouri's 74 non-users. So were 26.4% of those football players.