From Coin Flips to Earthquake Prediction: Understanding “Return Period” and “Once-in-a-Century Events”
From Coin Flips to Earthquake Prediction: Understanding “Return Period” and “Once-in-a-Century Events”
Wilson Chan Wai-soen
May 2026
We sometimes hear news reports about a “once-in-a-century” earthquake in an active region or
mentions of its “return
period”. These terms often cause confusion: does “once-in-a-century” actually mean that an
earthquake only happens once
every hundred years? To explain the true meaning of “return period” and “probability”, let’s begin
with a simple game of
coin flipping.
Imagine you are flipping a coin with a friend. Let’s use a one-dollar coin. One side with the
Bauhinia flower is “heads”
and the other side with the number is “tails”. In an ideal situation, the probabilities of flipping
heads or tails are
both half, i.e. 50%. Therefore, the return period for flipping heads is two tosses. In other words,
on average, you will
get one “heads” every two flips.
But what could happen if you flip the coin twice?
Figure 1 - A schematic tree diagram of coin flipping
Most of us might have experienced a similar game that, in most cases, you will get “heads” at least
once. However, it is
not hard to imagine that sometimes you indeed will not get “heads” at all in two tries. In an ideal
situation where each
flip is independent and does not affect the next, the probability of not getting “heads” in two
flips is one in four, or
25%.
“Return period” is one way of expressing probability. The higher the probability,
the shorter
the
return period; the
lower the probability, the longer the return period. However, a “return period” does not mean an
event is guaranteed to
happen within that fixed time period, nor does it mean that the event will not happen more than once
during that period.
Instead, it should be understood as an “average” expected value of the time period.
Now, let’s apply this concept to earthquakes.
An earthquake return period is an estimate of the average time interval between earthquakes of a
specific magnitude in a
particular region. Because the occurrence of earthquakes is random, current technology still cannot
precisely predict
when, where, or how strong an earthquake will be. To estimate the probability of a specific
magnitude earthquake
occurring in a given area, we look at historical seismic data spanning decades or centuries in that
region, count the
number of occurrences of that specific magnitude, and calculate the average time interval. Because
seismic activity
itself is highly variable and historical records may be incomplete, we can also fit historical
seismic data into
statistical models to estimate the earthquake return period.
Major earthquakes are the result of the accumulation and release of energy along faults. Due to the
relative motion of
tectonic plates, stress builds up in the slower-moving sections of the Earth's crust. When this
accumulated stress
exceeds the limit of friction between the two plates, the stored energy can be abruptly released,
causing an earthquake.
While the process of stress accumulation and release is cyclical, earthquake cycles are not entirely
regular.
Figure 2 - The repetitive process of stress accumulation and release
between
tectonic plates
Furthermore, an earthquake can sometimes trigger the release of accumulated stress in the vicinity,
serving as a
precursor to another earthquake event. Therefore, whether a major earthquake has occurred can
influence the likelihood
of another earthquake happening shortly after, in statistics, we say that earthquake events are not
independent. Let’s
take an example, say the chance of a magnitude 6 earthquake occurring in a certain region in a year
is 1%. We would call
this event has a return period of 100 years, or a “once-in-a-century” event. If we assume these
earthquakes in this
region are independent events (which is not actually the case, but can serve to illustrate the
concept), the probability
of no earthquake occurring within a century is about 37% (see Note below). This demonstrates that a
“once-in-a-century”
event does not mean the event must happen once within 100 years; the return period merely expresses
the probability of
occurrence. Commonly used return periods in seismology include approximately 475 years (a 10%
probability of exceedance
in 50 years) and around 2,475 years (a 2% probability of exceedance in 50 years).
Hong Kong is located within the Eurasian Plate, and is located rather far away from the nearest
plate boundary.
According to the study of earthquake hazards in Hong Kong by the Geotechnical Engineering Office of
the Civil
Engineering and Development Department, there are no obvious signs of recent fault activity in Hong
Kong or its
offshore, meaning the chance of a major earthquake is relatively low. However, people may travel to
or work in places
with higher seismic activity, it is essential to acquire disaster preparedness and prevention
knowledge and know how to
respond to hazards.
Note: Based on the multiplication rule for the probability of independent events, the detailed
calculation is given as
follows:
An event occurs on average once every 100 years (return period of 100 years), the probability of
occurrence in any given year is the reciprocal of the return period, i.e. 1/100 or 1%.
The probability of not occurring in one year is 100% − 1% = 99%
The probability of not occurring for two consecutive years is 99% × 99% = 98.01% (approximately 98%)
The probability of not occurring for 100 consecutive years is 99% × 99% × ⋯ × 99% (a total of 100 times), i.e. (99%)100 = 36.7% (approximately 37%)
The probability of not occurring in one year is 100% − 1% = 99%
The probability of not occurring for two consecutive years is 99% × 99% = 98.01% (approximately 98%)
The probability of not occurring for 100 consecutive years is 99% × 99% × ⋯ × 99% (a total of 100 times), i.e. (99%)100 = 36.7% (approximately 37%)