How often this is right
Measured two ways, reported separately, and published whether or not the number is flattering. Where there is not enough evidence to say, this page says that instead of printing a rate.
Against a labelled set
A fixed set of cases with known answers, which the product never learns from. This is the harder test: the cases are chosen to include the ones a person could not call either.
The set holds 0 scored cases, and 30 are needed before a rate means anything. A rate from a handful of cases is noise wearing a decimal point, so none is shown.
What the set is made of
Nothing yet.
Labelled by someone other than us: 0 of 0. Our own labels are worth less than an extension service’s, and this page says which is which rather than averaging them together.
From people who came back
Real lawns, where the person told us what it turned out to be. A different and messier question than the set above: these cases chose themselves, the labels are a homeowner’s rather than a laboratory’s, and only the people who came back are counted. It is reported separately for that reason.
| Diagnoses run | 19 |
| Photos kept, with permission | 15 |
| Outcomes reported back | 1 |
| Predictions answered | 0 |
| Cases that can be scored | 0 |
| Confirmed correct | not enough answers yet |
Nothing is scoreable yet. A kept photo is not a measurement: a case only counts once someone comes back and says what it turned out to be. 14 no outcome reported yet, 4 no permission to keep the photos, 1 the outcome was unresolved, so there is no label.
Nothing on this page is an estimate. Every figure is either measured or absent.