Fukushima tsunami and the hidden risk in backtests
Fukushima was built for the highest tsunami in 100 years of records, and the wave that came was 5 times higher. Your backtest has the same blind spot.
Education only, not investment advice. Figures here come from past data and backtests: illustrative, not live market data and not a performance claim.
On March 11, 2011, there was a massive earthquake in Japan. Tsunami waves started spreading towards the coast. Earthquakes and tsunamis are nothing new for the Japanese, but that day was different. What's the link between an earthquake in Japan and trading? There is one important common element: risk. Most traders don't even realise the risk they are exposed to. If it is not addressed, one day it could wipe out their trading capital. What is that risk?
A plant built on 100 years of data
Let's go back in time first. In the 1960s and 1970s, Japan was an upcoming major economy. Japanese brands like Toyota, Sony and Panasonic were being recognised globally, and this industrial expansion required immense amounts of electricity. The Fukushima nuclear plant, built in the 1960s and 1970s, was one such major supplier of electricity. It was located near the coast, in a region obviously prone to earthquakes and tsunamis. They knew about this while designing the plant.
In fact, Japan hired the best minds to design this plant. They rigorously studied 100 years of available geographic details about all the major earthquakes and tsunamis in that region and determined that a tsunami wave height of 3.1 meters was the maximum in the past. They built the nuclear plants between 10 and 13 meters above sea level, 3x to 4x what the historical study said. They were well prepared to handle these tsunami waves.
From 1970 to 2010, every time an earthquake happened, the Fukushima engineers didn't panic, because they were well prepared: they had studied the historical data and followed the protocols. Whenever such an earthquake happened, the protocols were followed automatically. The nuclear reactor would shut down, and the backup electricity in the basement would keep the reactors cool to avoid any kind of explosion due to overheating.
The day the wave crossed 15 meters
March 11, 2011 was a normal day at the Fukushima nuclear plant. Suddenly the earthquake alarm went off. As usual, the engineers didn't panic, until they saw tsunami waves at a staggering height of more than 15 meters. The engineers froze. They panicked and couldn't react, because according to their historical study, tsunami waves in that region had never crossed 3 meters. That day the wave was 5 times higher than anything that had happened in the past.
The water crossed the sea wall and entered the nuclear plant. It flooded the basement and disabled the diesel generators, leaving the plant without any electricity. About 24 hours after the tsunami, there was a hydrogen explosion. By 2013, the plant was decommissioned.
The Fukushima plant survived 40 years of 'normal' earthquakes. A nuclear plant that had worked perfectly fine for 40 years was totally devastated in its 41st year. The engineers at Fukushima were brilliant. They weren't careless. They just didn't respect tail risk.
They just didn't respect tail risk.
What your backtest cannot see
This is exactly what a trader does when he backtests a trading system. He tests 5, 10 or 15 years of data, and on that basis he thinks Nifty has never had a gap down of more than 5% in the past. "As per my backtest, my max drawdown is 20%, which was never breached in the last 15 years." "As per my backtest, my max losing streak was 5 back-to-back losses, so I'm well prepared." But let me tell you: a normal trend following or breakout system with 45% winning accuracy can have a losing streak of 18 back-to-back losses. Is your money management plan ready to handle these losing streaks?
Just because it has never happened in the past doesn't guarantee it will never happen in the future. Historically, Nifty has seen a drawdown of -66%. Does that mean that's the maximum risk one might face in equity investment? Not true.
The Dow Jones saw a drawdown of more than 90% in the 1929 crash, during the Great Depression. Our market data is just 25 years old, but Dow Jones data goes back more than 100 years.
The market doesn't care about probabilities. It doesn't care about your backtests. It doesn't care that something has 'never happened before.' The one who manages his risk well and is prepared for the worst-case scenario will survive long enough in this field.
The tsunami is coming. Maybe not today. Maybe not this year. But it's coming. Will your walls be high enough to handle it?
This article is based on a LinkedIn article by Groww that I read a long time ago.
I first wrote this as a thread on X. Follow me there for new ones.