On 7 August the count went 144, then 128, then back to 144 in a day. It looked like the site had gone backwards. It had not.
| What happened | Effect on the count | Was it progress? |
|---|---|---|
| 16 decorative placeholders removed — hero counts, section tallies, a metabar | −16 | Yes |
| 16 real ones added — a new out-of-sample block on three test articles, and a price-data row | +16 | Yes |
| Net | 0 | Both halves were |
A rising placeholder count means new slots a real test has to fill. A falling one is only good news when the numbers actually arrived.
When a test publishes a result. Two problems in our own data are holding that up, and both are recorded rather than worked around.
Because the method is the product as much as the numbers are. The Field Guide states no results at all and is finished, which is why it is the part open to search.
Not any more. That was the correction: a dash now only appears where a real test will put a number, and something on the page has to say so.
By running the command, not by remembering. That is the part that went wrong.
A placeholder may only appear in a slot a real test will fill, and something visible must explain it.
The count is taken by running the documented command every time, because a remembered figure once failed to reproduce and nobody noticed for weeks.
