Wilks and DOTS, worked down to the last coefficient
−1.291 × 10−8 reads like a typo. Multiply it by 905, which is 5,904,900,000, and that one coefficient strips 76.23 off a Wilks denominator of 783.22, roughly a tenth of the number a lifter's total gets divided by. The trailing digits are load-bearing, so this page prints them.
The formula: a bodyweight polynomial divides the total
Both systems have the same shape:
The 500 is a fixed numerator, identical for both. Everything separating the two lives in the denominator, a polynomial in bodyweight with a different coefficient set per sex.
The two polynomials are not the same degree. Wilks is a quintic: six coefficients, a through f, running from a constant up to bw5. DOTS is a quartic: five coefficients, a through e, stopping at bw4. Plenty of write-ups call both of them fifth-order. An early draft of this page did too, and we fixed it as a math error rather than a wording nit.
Coefflift stores both sets in ascending power order, so the array index is the exponent. The male Wilks set opens at −216.0475144 with a positive bw term of 16.2606339; the female set inverts that shape, opening at +594.31747775582 with a negative bw term of −27.23842536447.
Wilks and DOTS worked term by term, same lifter, same total
One lifter: male, 90 kg bodyweight, 600 kg total. Each row is one coefficient multiplied by 90 raised to that row's power.
| n | 90n | Wilks coefficient | Wilks term | DOTS coefficient | DOTS term |
|---|---|---|---|---|---|
| 0 (a) | 1 | −216.0475144 | −216.047514 | −307.75076 | −307.750760 |
| 1 (b) | 90 | 16.2606339 | 1463.457051 | 24.0900756 | 2168.106804 |
| 2 (c) | 8,100 | −0.002388645 | −19.348025 | −0.1918759221 | −1554.194969 |
| 3 (d) | 729,000 | −0.00113732 | −829.106280 | 0.0007391293 | 538.825260 |
| 4 (e) | 65,610,000 | 7.01863 × 10⁻⁶ | 460.492314 | −0.000001093 | −71.711730 |
| 5 (f) | 5,904,900,000 | −1.291 × 10⁻⁸ | −76.232259 | no term | n/a |
| denominator | 783.22 | 773.27 | |||
| 600 × 500 / denom | 383.0 | 388.0 | |||
Read down the two term columns and the systems look nothing alike. Wilks reaches its answer through a bw term partly cancelled by a −829.11 cubic, then rescued by a +460.49 quartic; DOTS gets there through a −1554.19 quadratic that dwarfs anything in the Wilks column. Different term counts, wildly different magnitudes, and they still land 9.94 apart on the denominator and 4.9 apart on the score. That closeness is an outcome of two independent curve fits, not of shared structure.
Where the coefficients come from, and why DOTS was refit
Coefflift uses the original Wilks formula, written by Robert Wilks of Powerlifting Australia in the mid-1990s. Its coefficients were fit against a competition dataset now more than two decades old and thin at the light and heavy tails, the origin of the accuracy complaint that follows Wilks around. Vanderburgh and Batterham later assessed it in Medicine & Science in Sports & Exercise 31(12) (1999) and found it valid, with a mild bias at those same tails. That paper tested the formula; it didn't rewrite the numbers. A genuine revision, Wilks Formula 2, arrived in 2020 with different coefficients. Coefflift does not use it.
DOTS is the official 2019 release, refit against a larger, more recent pool of results and adopted by several federations as the modern replacement. It also dropped the fifth-order term entirely, a modeling decision rather than a presentational shortcut.
We don't print any federation's current rulebook here. Rulebooks change season to season, and a stale claim on this page would be worse than no claim at all.
Rounding rules, plausibility bounds, and where the polynomial breaks
The arithmetic runs at full float precision on canonical kilograms. Pounds convert at 1 kg = 2.2046226218 lb. The conversion happens before the polynomial and never after, so both scores are unit-invariant: toggling kg/lb changes the displayed inputs and leaves the scores identical. Rounding to one decimal is display only and never feeds a later step. The worked numbers above are rendered through the same lib/calculator entry point the live scoreboard calls, so a published example can't drift from the tool that produced it. We chose that over hand-transcribing the values into prose, which is how most wrong worked examples get written.
The bounds exist because the polynomial genuinely breaks. Bodyweight is accepted in [25, 300] kg, totals up to 2000 kg. Both denominators stay strictly positive from 30 to 250 kg, and the Wilks quintic's denominator only turns negative past roughly 300 kg, where the score would flip sign and read as a plausible-looking negative number. Capping at 300 kg means the math never runs on a negative denominator. The 2000 kg cap sits well above the ~1400 kg raw world-record three-lift totals, so anything past it is a data-entry error rather than a lift.
Inside the plausible range sits a second, softer band. Below 45 kg or above 180 kg both scores still render, plus a caveat that Wilks is least reliable out there. The 90 kg lifter in the table scores with no caveat at all.
Blank lifts in per-lift mode count as 0 kg rather than blocking the sum. All three blank collapses the scoreboard back to its dash, never to 0.0, because a zero score is a claim and an empty field is not.
Frequently asked questions
More gym math
