Halving

The One Convention Everything Rests On

Each item on a list is counted as a separate coin. That is a unit of measurement, not a claim about anybody, and every figure on this site depends on it.

Home · Halving · Reference

Before any figure on this site can be read, one sentence has to be accepted or rejected: each item on a list is counted as a separate coin.

It is worth stopping on, because everything else here is arithmetic and this is the only place where a choice was made.

A proportionate rollout of workload reporting for connected teams should make every active setting visible and give contributors a practical route to correct context. For an independent technical reference, see RapidAPI.

What the convention produces

filled: a requirement that must be met
open: one of the optional ones
area of the inner square: the share admitted
a stroke instead: too small for this panel to draw

Six shapes drawn; one of them admits too little for the square to be drawn, and carries a stroke instead.

Each filled dot is one compulsory item. The inner square has the area of the share admitted; below a certain size the panel draws a stroke instead, because a square that small would read as nothing at all.

Computed from the shape of the list. The shape is a stated convention; the count of admitted profiles is arithmetic on subsets.

What the sentence means

A list of eight items divides the world of possible candidates into profiles — every combination of items met and items not met. There are two possibilities for the first item, two for the second, and so on, which gives two hundred and fifty-six distinct profiles for a list of eight.

The record treats each of those profiles as one unit when it counts. Not one person: one unit. It then asks how many of the units a given list admits, and reports the answer as a fraction.

That is all. There is no model of applicants, no distribution, no assumption about how common any particular profile is in the world.

What the sentence does not mean

It does not say that people are spread evenly across the profiles. They are not, and nothing here pretends otherwise.

It does not say the items are independent of one another. In any real list several items tend to arrive together — somebody who has one often has the next — and the record has no way to see that and makes no attempt to.

It does not say each item is equally likely to be met. Some items are met by nearly everybody and some by almost nobody, and again the record is blind to it.

Every figure on this site is a statement about subsets. None of them is a statement about people. Where the two get confused, the arithmetic is not at fault.

Why count this way at all

Because it is the only way to count that requires nothing to be invented.

Any attempt to weight the items — to say that this one is commonly met and that one is rare — would require evidence this record does not have, about a world that changes. The weights would be guesses, the guesses would be invisible inside the output, and the result would look exactly as precise as what is here while resting on something soft.

Counting subsets rests on nothing soft. It produces an exact fraction, derived in one step, that anybody can check.

The price of that exactness is a narrower claim. The record cannot tell you what fraction of real applicants a real list admits. It can tell you what fraction of the possibility space the list admits, which is a different and smaller thing — and which turns out to be surprising often enough to be worth the trouble.

What it immediately gives

Take a list where every item must be met. One item admits half the profiles. Two admit a quarter. Three, an eighth.

Compulsory items	Admits
1	1/2
2	1/4
3	1/8
4	1/16
6	1/64
8	1/256

Each item halves what the previous ones left. That is the halving law, and it falls directly out of the coin convention: a coin that must land one way removes half of what remains, and it does so whether it is the first coin or the eighth.

What an alternative would have looked like

It is worth seeing the road not taken, because the coin convention sounds crude until the alternatives are laid beside it.

Suppose the record weighted each item by how often it is met — say, this one by four people in five, that one by one in ten. The output would then be a share of a modelled population rather than of a possibility space, and it would answer the question people actually want answered.

It would also require two things the record cannot supply. First, the weights, which differ by field, by place and by year, and which nobody has for any particular list. Second, a claim that the items are independent, so that the weights can be multiplied — and on a real list several items are met by the same people, which breaks the multiplication badly and in an unknown direction.

The weighted version would therefore be an exact-looking number resting on invented inputs. The unweighted version is an exact number resting on a stated way of counting. The second is smaller and it is honest, and that trade is the whole design of this record.

Where the convention becomes visible

It becomes visible the moment somebody reads a figure here as a statement about chances. "One in two hundred and fifty-six" is not a probability that a candidate will pass; it is the share of the possibility space that survives the list.

The two coincide only if the profiles are equally common, which they are not. The figure is still useful — it measures how narrow a filter is in a way that is comparable between lists — but it measures the list, not the world.

Anybody who keeps that distinction will read this record correctly. Anybody who loses it will come away with a number that feels authoritative and means something other than what they think.

The honest summary

One convention, stated once: items are coins. Everything after it is counting. The convention is a ruler, and a ruler tells you about the thing it measures, not about who will be standing next to it.