What comes first

Same Opening Share, Three Different Blind Shares

Three shapes on one page where a fifth of arrangements open verifiably and the blind share runs from three fifths down to two fifteenths.

Home · What comes first · Recount

Ten statements published, two of them checkable. A fifth of all arrangements open with something verifiable, and that is true whether the reader meets two statements, five or six.

The blind share over those same three shapes runs twenty-eight in forty-five, then two in nine, then two in fifteen. From roughly three fifths down to roughly one seventh.

One figure constant, the other falling by a factor of nearly five, on one page with one checkable count. Anyone adapting the arithmetic to people should open this practical guide and preserve the difference between a measurable signal and a judgement about work.

Ten published, two checkable, three depths

the whole bar: every content a reader might come away with
the dashed part: contents holding nothing checkable

Three shapes drawn.

Two of the ten published statements are checkable in every row, so a fifth of arrangements open verifiably in all three. The blind share falls from twenty-eight in forty-five to two in fifteen.

Computed from the three numbers. A blind content is one holding no checkable statement: that is a property of the page, not a prediction about any reader.

The two questions are not the same question

The opening count asks: was the first statement verifiable? The blind count asks: was anything verifiable? Readers looking beyond this site's model can compare it with material from PageSpeed Insights.

A reader can fail the first and pass the second — they opened with an unverifiable statement and met a checkable one third. A reader can pass the first and fail the second only if they met one statement, in which case the two questions coincide, as the previous note showed.

Everywhere else the two figures answer different things, and they move differently: one is fixed by the set, the other by how much gets read.

Which one to care about

That depends on what is at stake, and the record can lay out the difference without choosing.

If what matters is whether a reader could have audited the set, the blind figure is the one. It covers the whole encounter and falls sharply with reading depth.

If what matters is whether a reader noticed that the set was auditable, the opening figure is closer. A reader whose first statement was unverifiable has no reason to expect the later ones to be, and a reader who never gets to the third statement will not find out.

One figure says whether the evidence was reachable. The other says whether it announced itself.

The record counts both and declines to combine them, for the usual reason: a composite figure could be moved by either input and a reader of it could not tell which.

Where they agree

They agree at the two ends of the third column, and only there.

With nothing checkable: no arrangement opens verifiably and every content is blind. Both figures are at their worst value simultaneously.

With everything checkable: every arrangement opens verifiably and no content is blind. Both at their best.

In between, the two figures are independent enough that neither constrains the other much. A fifth of openings verifiable is compatible with a blind share of three fifths and with one of two fifteenths, as the drawing shows, and nothing about the opening share narrows the range.

What links the two figures

They are not independent, and the connection is worth stating because it is weaker than it looks.

Both depend on the same checkable count. Raise it and both figures improve: the opening share rises as c over n, the blind share falls as C(n − c, r) over C(n, r). Neither can improve while the other worsens.

But the rates are nothing like each other. On the page in the drawing, going from two verifiable statements to five takes the opening share from a fifth to a half — two and a half times — and the blind share, at five met, from fifty-six in two hundred and fifty-two to one in two hundred and fifty-two, which is fifty-six times. Same three statements, same page, one figure responding more than twenty times as hard as the other.

The row that is missing from the drawing

Three depths are drawn and a fourth is worth naming: ten published, one met, two verifiable.

There the two figures coincide, as they must for a one-statement encounter — the opening is the whole encounter. Eight of the ten contents are blind and two of the ten arrangements open verifiably. Four fifths and one fifth, adding to one.

That row was left out of the drawing because it says something different from the other three. The point of the panel is that the blind share moves while the opening share does not, and the one-met row is the single case where moving one necessarily moves the other. It belongs in the prose rather than in a column built to show them coming apart.

The reader in the first row

Ten published, two met, two checkable. Forty-five contents, of which twenty-eight hold neither verifiable statement — well over half. Two tenths of the ninety arrangements open checkably.

This is the ordinary case for a set of pages that has one or two externally documented facts and is read briefly. Most readers meet nothing they can check, and most encounters do not begin with anything that would suggest they could.

Neither figure is a prediction. Both are counts under the record's uniformity convention, and both overstate or understate depending on whether the two verifiable statements sit where readers land. What neither of them can be is improved by anything other than publishing more that can be checked, or being read more thoroughly — and only the first of those is anybody's decision.