Making weight

Abstract

We study 536,668 raw full-power results from 10,585 meets with IPF weight classes since 2012. Bodyweights pile up just under every class limit: 5 to 15 times as many men weigh in up to 1 kg under a limit as up to 1 kg over it, and 7 to 10 times as many women. At the same weights in meets with traditional classes, where no limit lies there, the ratio is below 1. Lifters just under a limit total far more than lifters just over it, but that comparison is selection: whoever cuts tends to be the more serious lifter. Comparing the same lifters at the same limit, the meets where they made the lighter class were about 2.1 kg lighter, with practically the same total (−0.1% for men, −0.5% for women), and a rank in the class +11 points higher for men and +8 points for women.

In plain words

13×
as many men weigh in just under the 83 kg limit as just over it. The limit, not the body, decides where lifters end up.
−0.1%
is the total of the same men when they made the lighter class, against meets where they weighed in just over it, about 2 kg heavier. The cut costs next to nothing on the platform.
+11 points
is what that buys in the class ranking. 84% of these men rank higher in the lighter class.

Data

The data is the public OpenPowerlifting archive. We keep every valid raw full-power result (squat, bench, deadlift) from a sanctioned meet since 2012, leaving out high-school divisions. A lifter entered in two divisions of the same meet appears once, with the higher total.

What makes a weight a limit is the class system of the meet. We read it from the classes in the results: a meet uses IPF classes if every class limit in it belongs to the IPF set (for men 53, 59, 66, 74, 83, 93, 105 and 120 kg) and at least three appear. Meets whose limits all belong to the traditional set (for men 52, 56, 60, 67.5, 75, 82.5, 90, 100, 110, 125 and 140 kg) serve as a cross-check.

The study looks at bodyweights at the official weigh-in. How much a lifter weighs on a normal day is not in the data.

The pile-up

Take every entry in a meet with IPF classes, and measure the bodyweight from the nearest class limit.

Bodyweights pile up just under the class limits

Figure 1. Bodyweight minus the nearest class limit, meets with IPF classes, in steps of 0.25 kg. Red: up to 1 kg under the limit. A lifter at exactly the limit makes the class.

Third Attempt by Daniel Vahl · Data: OpenPowerlifting, current dump

The picture is a saw: entries rise toward each limit, then collapse right after it. A simple measure of the jump is the ratio of entries just under a limit to entries just over it.

The math: the pile-up ratio

For a limit $L$ and bodyweights $w$ , let $N(a < w \le b)$ count the entries in that range. The pile-up ratio is

R_L = \frac{N(L - 1 < w \le L)}{N(L < w \le L + 1)}. \tag{1}

Without a limit at $L$ , and with a smooth distribution of bodyweights, $R_L$ would be close to 1.

Men weigh in just under a limit 5 to 15 times as often as just over it, least so at 120 kg, where fewer lifters still cut. Women do so 7 to 10 times as often.

Is it the limit? Bodyweights could bunch at certain values for other reasons, such as rounding on the scales. The cross-check settles it: at the same weights, in meets with traditional classes where no limit lies there, the ratio is 0.6 to 1.0 for men and about 0.8 for women: no pile-up at all. That it lies somewhat below 1 is expected, because in those meets the bodyweights rise toward the next traditional limit.

Pile-up at each limit, with and without a limit there

Figure 2. Pile-up ratio (1) at each IPF limit, in meets with IPF classes and in meets with traditional classes. “No test” where a traditional limit lies within 1 kg and would bunch there itself.

Third Attempt by Daniel Vahl · Data: OpenPowerlifting, current dump

Does the cut cost strength?

The tempting comparison. Men who weighed in up to 1 kg under a limit have a median total of 550 kg; men up to 1 kg over it, 495 kg. That is +11% for the lighter group (+13% for women). It looks like a regression discontinuity: two groups on either side of a sharp threshold, as similar as can be except for the class.

The math: why the discontinuity design fails here

A regression discontinuity compares units just below and just above a threshold in a running variable, here bodyweight. It is valid only if units cannot sort themselves precisely around the threshold, so that those just below and just above are alike. Figure 1 is the test for exactly this, the density test of McCrary: the density of the running variable must be continuous at the threshold. Instead it jumps: for men there are 5 to 15 times as many entries just under a limit as just over it. Lifters choose their side of the limit, and the ones who cut are not the ones who don’t.

The lifters just over a limit are mostly those who did not try to make it: newer, less competitive lifters, or lifters who simply compete at whatever they weigh. The gap says who they are, not what the cut does.

The same lifters. Following each lifter over their own meets removes who they are. Two measures do this.

The math: the within-lifter model

Let $T_{it}$ be the total of lifter $i$ at meet $t$ , $w_{it}$ the bodyweight, $L_{it}$ the limit of the class they made, $Y_{it}$ the years since their first meet and $K_{it}$ the number of the meet, capped at 6. With a lifter effect $\alpha_i$ ,

\log T_{it} = \alpha_i + \beta\, w_{it} + \gamma\, \mathbf{1}[0 \le L_{it} - w_{it} \le 1] + \delta\, Y_{it} + \sum_{k=2}^{6} \kappa_k \mathbf{1}[K_{it} = k] + \varepsilon_{it}. \tag{2}

$\beta$ is what a kilogram of bodyweight is worth for the same lifter, $\gamma$ what weighing in just under the limit is worth at the same bodyweight. Intervals come from 200 bootstrap samples of whole careers.

For lifters who made a limit $L$ at some meets (up to 1 kg under) and missed it at others (up to 3 kg over), the pair contrast is

\Delta_i = \overline{\log T_i}\,\big|_{\text{under } L} - \overline{\log T_i}\,\big|_{\text{over } L}. \tag{3}

For the same man, a kilogram of bodyweight is worth about 0.73% of total, for a woman 0.57%. Being just under the limit, at the same bodyweight, is worth +1.2% for men (95% interval +1.1% to +1.4%) and +0.6% for women. If cutting cost strength, $\gamma$ would be negative. It is positive.

The pairs (3) say the same in kilograms: 6,069 men and 5,854 women made a limit at some meets and weighed in just over it at others. Where they made it, they were about 2.1 kg lighter and totalled practically the same (Table 2). Had those kilograms been lasting bodyweight, $\beta$ from (2) would predict about 1.5% less for the men.

**Table 2.** Under against over the limit, three ways. Intervals: 95% bootstrap.
	Men	Women

A kilogram of lasting bodyweight costs strength, a kilogram cut for the weigh-in barely does. Most of a short cut is water and food, which come back after the weigh-in.

This does not prove that cutting is free. Lifters cut most for the meets they care about, and they peak for those meets, too. That selection pushes $\gamma$ up. What the data does show: any cost of the cut must be smaller than the boost from peaking that hides it.

Does it pay?

The cut buys a lighter class. For the pairs from (3) we take each entry’s rank among all entries in that class across the archive, from 0 (lowest total) to 100 (highest). The open class above the top limit counts as a class of its own.

Rank in the class, over the limit against under it

Figure 3. Average rank in the class for the same lifters, at meets where they weighed in just over a limit (class above) and where they made it (class below).

Third Attempt by Daniel Vahl · Data: OpenPowerlifting, current dump

Women gain +8 points on average and men +11 points. With the same total, the lifter is now among lighter competitors. 84% of the men rank higher in the lighter class.

Limits and next steps

Weigh-in timing. Federations differ in how long before lifting the weigh-in happens. The archive does not record it, and the cost of a cut probably depends on it.
Failed cuts. A lifter who tried to make 83 kg and missed shows up as just over the limit. Some of the “over” entries are failed cuts, which would make missing a limit look worse than not trying.
Health. The data only shows lifters who made it to the platform. It says nothing about the risks of cutting, which grow with the size of the cut.

How it was computed. The dataset is built from the OpenPowerlifting dump with DuckDB, and the models are fitted in Python with NumPy. Every number on this page is recomputed with each data update.

Should you cut?

The calculator at the end of the page shows what a cut would do to your rank. Enter your bodyweight on a normal day and your total. The calculator takes the next class down, assumes you keep your total, as the lifters in (3) did, and ranks you there among everyone in the calculator on the home page. The note shows the rank if the cut cost strength like lasting bodyweight, with $\beta$ from (2):

T_{\text{after}} = T \, e^{-\beta c}, \tag{4}

where $c$ is the cut in kilograms.

References

OpenPowerlifting. OpenPowerlifting data dump, public domain. openpowerlifting.org.
McCrary, J. (2008). Manipulation of the running variable in the regression discontinuity design: a density test. Journal of Econometrics 142, 698–714.
Lee, D. S. and Lemieux, T. (2010). Regression discontinuity designs in economics. Journal of Economic Literature 48, 281–355.
Saez, E. (2010). Do taxpayers bunch at kink points? American Economic Journal: Economic Policy 2, 180–212.
Kleven, H. J. (2016). Bunching. Annual Review of Economics 8, 435–464.
Wooldridge, J. M. (2010). Econometric Analysis of Cross Section and Panel Data, 2nd ed. MIT Press.