Home › The Moon Test, explained
The Moon Test, explained
One question, asked in advance: is your catch per hour higher inside solunar windows than outside? The test answers with a ratio and an honest interval.
The windows
| Window | Definition | Length |
|---|---|---|
| Major | Moon overhead or underfoot (meridian crossing) ±1 hour | 2 hours |
| Minor | Moonrise or moonset ±30 minutes | 1 hour |
Positions come from SunCalc and were checked against an independent astronomical ephemeris (ephem) in several cities; centres agree within a few minutes, far less than the window width.
The measurement
- For each finished fishing trip, count the hours inside windows and outside, and the catches in each.
- Add up across all trips: catches per hour inside, catches per hour outside.
- The ratio is inside divided by outside. 1.0× means no difference; 2.0× means twice as many fish per hour inside the windows. A half-catch is added to each count so the ratio is defined even when a count is zero.
The interval
Moonlog resamples your trips with replacement 2,000 times (a trip-level bootstrap, with a fixed seed so the result is repeatable) and reports the 2.5th to 97.5th percentile as a 95% interval. Resampling whole trips rather than single fish respects the fact that fish caught on one day are not independent.
The verdict
| Trips | Interval | Verdict |
|---|---|---|
| Fewer than 10 | any | Too few trips to conclude anything |
| 10 or more | entirely above 1.0× | Higher catch rate inside the windows |
| 10 or more | entirely below 1.0× | Lower catch rate inside the windows |
| 10 or more | includes 1.0× | No detectable difference |
"No detectable difference" does not mean there is none: with few trips a real effect can hide in a wide interval.
How well does it work?
We tested the method on simulated trips with a known truth (30 trips at 0.6 fish per hour). With no effect, the 95% interval contained the true ratio of 1.0 in about 95% of simulations, and a wrong "higher" or "lower" verdict appeared about 5% of the time, as expected. With a true doubling inside windows, the test said "higher" in about 95% of simulated datasets. Real data is messier than a simulation.
Limits you should know
- Dawn and dusk. Windows often overlap them. A real dawn effect can look like a moon effect.
- One hypothesis. The test asks one question set in advance. Slicing by species, season or place makes false positives more likely, so Moonlog labels any other breakdown exploratory.
- Effort and conditions. Weather, fishing pressure, bait and where you fish all matter and are not controlled.
- Self-selection. If you only go out in windows, there is nothing to compare.
Open Insights The app has sample data you can load to see how a clear result looks. It is labelled as sample data.
Frequently asked questions
How many trips do I need?
About 30 for a first read. The interval narrows as trips are added.
Does it work for hunting?
The Moon Test uses fishing trips. Hunting sightings feed the wind rose instead.