The Predicted Quake That Did Not Happen

1. Post hoc pattern rather than a tested prediction. The club looked at events that already happened and then searched for a lunar connection. With 12–13 full moons each year, some earthquakes will fall near a full moon by chance. A scientific prediction states in advance what will happen, what would count as failure, and how many events will be tested.

2. Ignoring the base rate. The catalog shows 19 M3+ events in 25 years but no M6+ in 60 years. That makes a M6+ in any specific seven-day window very unlikely. The claim shifts attention from the ordinary explanation-small earthquakes are common-to an extraordinary one without extraordinary evidence.

3. Selection and detection bias. Using only nine events from the past decade is not a complete sample. The 2012 station additions also changed what the network can detect, so apparent clustering may reflect better recording, not real change.

4. No quantitative mechanism. Tidal stress is real, but its stress change is orders of magnitude smaller than tectonic stress. Saying tides 'trigger' quakes is not a mechanism unless the club specifies the stress threshold, depth range, and expected timing.

5. Unreliable source and no peer scrutiny. A blog post with a red circle is not a primary analysis. The claim has not been independently checked, replicated, or reviewed.

The claim is phrased in a falsifiable way, which is necessary but not sufficient. It fails because it was generated after the fact, ignores base rates, uses a biased sample, and supplies no testable physical model.

Students should write the protocol before looking at the data. The protocol should define:

• Region: a polygon around the school, not a circle drawn after an event.
• Time window: the full 25-year catalog, including years before 2012, with a note about detection changes.
• Magnitude cutoff: for example, all M3.0+ events, chosen in advance.
• Lunar variable: phase angle or days from full moon, binned into equal intervals.
• Test statistic: observed counts per bin compared with the counts expected if earthquakes were unrelated to phase.

Students then use every qualifying event, not a hand-picked subset. They compare the observed distribution with a uniform expectation and discuss whether any difference is larger than what chance and catalog incompleteness could produce. They should also check a second region or a later time period as a replication test. The key protections are pre-registration of the rule, use of the full catalog, an explicit null expectation, and replication. If the result is 'no association,' that is a valid scientific finding, not a failure. The activity also lets the teacher distinguish a testable hypothesis from a prediction made after the data are known.

Draft message: 'We understand that families are concerned about a social media post predicting an earthquake this week. The district has reviewed the claim with the available seismic catalog. We found no validated method that can predict the time, location, and size of an earthquake within a seven-day window. The post used a small, selected set of past earthquakes and did not test its rule in advance. School will remain open on the normal schedule. The most useful action is to practice Drop, Cover, and Hold On and to follow official emergency alerts. We will share any verified information from state emergency officials.'

Why this works: It names the specific weakness-no validated short-term prediction-without ridiculing the students or implying that earthquakes are impossible. It separates 'we cannot predict this event' from 'earthquakes are a real hazard.' It gives a concrete preparedness action, which reduces anxiety and replaces rumor with a procedure. It does not promise safety or give a false all-clear. It also avoids repeating the red-circle map or the blog's framing, so the school does not amplify the original claim.