Thread by @celestelabedz

How do earthquake “prediction” con artists make it LOOK like they have a good track record, even though they’re totally unscientific hoaxters?

Let a seismologist fill you in.

Thread:

First, to be abundantly clear: no one can usefully predict earthquakes before they happen. Not you, not your pet, not some guy on the internet. We’d all love if good predictions were possible (seismologists included!), but nothing yet has stood up to scientific rigor.

Why talk about this? With the recent swarm on the Brawley Seismic Zone, we (again) saw prediction charlatans try to prey on anxieties and peddle misinformation. That sucks, of course, but it’s also actually dangerous, because it can muddle important information from real experts.

These hoaxters are skilled at their craft, though. If you go to their web sites, YouTube channels, or wherever they’re trying to get clicks/money from you, you’ll find some time-stamped predictions and reports of the quakes that followed. At first glance, they can look good.

You might think “well, they just make tons of predictions, and delete all the ones that didn’t work out,” and yeah, they do that, too. But con artists know you can actually do better than that in the first place with a simple trick:

Be LIKELY, but not USEFUL.

How can a prediction be likely, but not useful? First, let’s define what a “prediction” is for an earthquake, because it matters. A prediction includes three things: time, location, and magnitude ranges for a quake that hasn't happened yet.

So, a prediction could take a form like “I predict there will be an earthquake with magnitude greater than 5, in the U.S. state of Alaska, in the next month.”

Then we wait. If it comes true, then wow!! I’m on to something, right?! Not quite.

The trick: that prediction is likely to come true, no matter when I say it!

Alaska is BIG and has LOTS of earthquakes, especially on the Aleutian Islands (because that’s a subduction zone). Here’s a map of all of the magnitude >5 events in the last two years; there’s over 100!

Those hoaxters are counting on the fact that the average person doesn’t know that earthquakes like that happen so frequently. And it’s okay if you don’t! Most of those quakes don’t make the news, since few (if any) people feel them, so you're unlikely know unless you look it up.

A magnitude 5 quake can sound scary. If one was going to happen below your house, you’d love to know what month it would happen in! Con artists count on you conflating a useful prediction (a quake near you) with an un-useful prediction (one in a huge, earthquake-prone state).

It’s like predicting “there will be rain this week”. If you say that on a random day in Seattle, you’ve got good odds, but being right is easy enough that it doesn’t mean much. To someone who lives in Phoenix, however, that can sound like some sweet prediction power!

There’s a wide variety of this kind of prediction that charlatans can make to pad their records in hopes of fooling you.

Magnitude >6 somewhere in the world in the next few weeks?
Likely, but not useful.

Magnitude >2 in southern California today?
Likely, but not useful.

So how do you check (and bust!) a hoaxter’s claim? The @USGS has your back with their earthquake catalog, searchable by place, time, and magnitude, so you can see how often quakes REALLY happen: earthquake.usgs.gov/earthquakes/se…

Important reminder: predictions are NOT the same as forecasts (sets of possible quake scenarios and their probabilities) or earthquake early warning (telling you existing seismic waves are heading your way). Research-backed aftershock forecasts & early warning are real and cool!

Two days after posting this thread, my example of a non-useful prediction came true: earthquake.usgs.gov/earthquakes/ev…

That's what happens when you predict things that are very likely! If you're blown away by my skill, then get this: I predict that the sun will rise tomorrow morning. 😉