How long a panel is

Every panel prose is flattened into one string, tags are stripped, whitespace is collapsed and the result is split on spaces. That gives a word count per panel, and the three tiles reduce that set to its average, its minimum and its maximum.

The fourth tile plots every panel on the site in order, so the range is not just two endpoints. The last tile gives this panel share of the whole board, which is a quiet way of saying that no single panel is a large part of it.

Because a target was set before anything was written, and panels were expanded or trimmed to land inside it. The spread you are looking at is the spread of a rule being followed, not the natural length of these subjects.

This is the same point as on the section sizes panel, and it is worth making twice. Regularity in a set of numbers looks like evidence. Here it is evidence of a decision, and the only reason you can tell is that the panel says so.

The shortest panel is the one where the material ran out earliest, and the longest is the one where it did not. Those are mildly interesting and much less informative than they look.

A short panel can be a subject with little to say, or a subject this board has no data about, or a panel that was written tightly. A long one can be a rich subject or a loose one. Nothing in a word count separates any of that, which is why the tiles are here alongside a paragraph saying so.

Because the alternative is a site that has an unstated house length nobody can see. Publishing the range makes the rule visible, and a visible rule can be judged.

It also does something to the writing. A rule you have published is a rule you notice yourself breaking, which is a mild but real discipline, and the corpus section exists partly to impose that on the rest of the board.

• Whether any panel is worth reading. Length is silent on that.
• Whether a panel is padded. A padded panel and a dense one of the same length are indistinguishable by count.
• Whether the target was the right one. That is a judgement about readers, and this board has no data about readers.
• How much of a panel anybody reads, which is unknowable here and always will be.

This panel, the one before it and the one before that all reduce to the same admission: the numbers are exact and the things they measure are weak proxies for anything anybody cares about.

That admission is the point of the corpus section rather than a weakness in it. Every other section of this board asks you to accept a number without being able to check it against the world. Here you can check every figure completely, and what you find when you do is that exact numbers about a text tell you very little about the text. That is worth knowing before you look at any other dashboard.

Two kinds of panel had trouble with the target length, and they had trouble in opposite directions.

The string panels could have run much longer. There is a great deal that can be computed about three strings, and most of it is uninteresting, so the writing there was mostly a matter of leaving things out. Those panels are trimmed rather than filled.

The money and orders panels are the other case. There is no data behind them at all, so the material is structural reasoning rather than measurement, and structural reasoning runs out faster than arithmetic does. Those panels were the ones that needed working at, and if you find any of them thinner than the rest, that is where the thinness came from.