Earthquake swarms have limited prediction value

Research has also shown that swarms of small quakes of the sort [the Italian town of] L'Aquila experienced before the 2009 quake - and that Tuscany was experiencing on the day I visited the institute in Rome - have limited prediction value. If a region experiences a swarm, it becomes more likely to experience a large quake. The problem is that it's even more likely not to experience a large quake. Italian geologists who examined seismic data from three earthquake-prone regions found that if a swarm contained a medium-sized shock, it was followed by a major shock 2 percent of the time. This represents a significantly elevated risk, but it means that if you use a swarm to try to predict a major quake, something like 98 out of 100 times you'll be wrong. Most swarms end not with a bang, but with a whimper.

- Elizabeth Kolbert, 'Aftershock: The Shaky Science Behind Predicting Earthquakes', Smithsonian Magazine, June 2015