Nothing to check

What One More Verifiable Statement Buys

Each verifiable statement added multiplies the blind count by a harsher fraction. The first addition helps least, the last one ends the question.

Home · Nothing to check · Recount

Ten statements published, a reader meets four. Watch the blind count as verifiable statements are added one at a time.

Nothing checkable: two hundred and ten blind contents, which is all of them. One checkable: one hundred and twenty-six. Two: seventy. Three: thirty-five. A reviewer who wants to connect the abstract count with day-to-day work can see the official platform overview before designing a limited pilot.

The steps are eighty-four, fifty-six and thirty-five. Each addition removes fewer blind contents than the one before, and yet each one is a larger proportional cut. Readers looking beyond this site's model can compare it with material from MDN Web Docs.

Ten published, four met, the count climbing

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

Four shapes drawn; one is blind whatever the reader meets.

One page, four shapes. The checkable count goes nought, one, two, three and the blind count falls two hundred and ten, one hundred and twenty-six, seventy, thirty-five.

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 column read from either end

In absolute terms the returns diminish. The first verifiable statement removes eighty-four blind contents; the third removes thirty-five.

In proportional terms they accelerate. Two hundred and ten to one hundred and twenty-six is a cut of two fifths. One hundred and twenty-six to seventy is a cut of four ninths. Seventy to thirty-five is a cut of one half.

Both readings are correct and they point opposite ways, which is why the record gives the counts and the contents figure rather than a single summary of improvement. The honest statement is that each statement added removes a smaller slice of a smaller remainder, and the slice is a larger fraction of what is left.

The factor, written out

Going from c checkable statements to c + 1 multiplies the blind count by (n − c − r) over (n − c).

On this page, with ten published and four met: from nought to one the factor is six tenths; from one to two it is five ninths; from two to three it is one half; from three to four it is three sevenths.

step	blind count	removed	factor
none to one	210 → 126	84	six tenths
one to two	126 → 70	56	five ninths
two to three	70 → 35	35	one half
three to four	35 → 15	20	three sevenths

The numerator falls by one each time and so does the denominator, so the fraction drifts downward and keeps doing so until the numerator reaches nought. That happens when the unchecked count equals the met count, and the step after it takes the blind count to nothing.

The last step is the big one

On this page the threshold arrives at seven checkable statements. Six checkable leaves four unchecked against four met: one blind content. Seven checkable leaves three unchecked, fewer than four met, so the blind count is nought.

So the final addition takes the count from one to nought — a factor of nothing at all — while the first took it from two hundred and ten to one hundred and twenty-six. Measured by what it removes, the first statement is worth eighty-four times the last. Measured by what it achieves, the last one ends the question and the first one does not.

The statement that closes a gap is worth more than the statement that narrows it, and no count of blind contents can express that. The threshold is a different kind of fact from the counts either side of it.

What this does not license

It is tempting to turn this into guidance: verify the last few, they pay best. The record cannot support that, for two reasons.

The marginal factors above assume every content is equally reachable, which is the second declared convention and is not true. If the verifiable statements are the ones readers rarely reach, none of these factors describes anything that happens.

And the record knows how many statements are checkable, never which. Adding a verifiable statement in the record means incrementing a count. Doing it in practice means writing a particular sentence about a particular externally documented fact, and whether that sentence is worth writing depends on what it says — which is outside this record permanently.

The step before the threshold

Two shapes on this page are worth setting side by side: six checkable and seven.

At six, four statements are unchecked against four met, and there is exactly one blind content out of two hundred and ten. At seven, three are unchecked, fewer than four met, and there are none.

A reader looking at the first of those sees a share of one in two hundred and ten and would reasonably call the question settled. It is not settled; it is small. The difference between small and settled is the difference between a figure that depends on the record's uniformity convention and a figure that does not, and on this page it costs one statement.

The page beneath the drawing

All four rows have two hundred and ten contents and five thousand and forty arrangements. Nothing in the first two parts of the record distinguishes them.

A reader handed only those two figures would see one page. The blind count across the four drawn shapes runs from all of it to one sixth of it, and reaches nought four steps further on. That is the whole argument for carrying a third number, made on a single page.