“The inability to predict outliers implies the inability to predict the course of history.”
I had been up for hours. My tired eyes strained against the bright glare of my computer monitor. Not again. As I stared helplessly at my profits and losses, the red arrows were unrelenting. This is the third night in a row. At 18-years-old, I already fashioned myself an amateur hedge-fund manager. With the help of my physics teacher, I had formulated and applied a number of quantitative trading algorithms to the local stock market, the Karachi Stock Exchange. It hadn’t worked out too well, but I was determined to improve my models till they could predict the market more accurately.
As I switched off the screen, rubbing my eyes and stretching my tense muscles, I realised that what I was actually predicting was human behaviour. What could be harder? Had I been fooled by randomness? Did I see patterns where none existed?
Many quantitative approaches to trading failed to recognise what Professor Nassim Taleb calls black swan events during the stock market crash a few years ago. Sophisticated models based on a dizzying array of differential calculus, advanced geometry and formulas borrowed from physics could not predict the most elusive variable mathematics has ever been confronted with: human behaviour.
As a quantitative thinker and trader, this inability to model human behaviour often frustrates me. But as I delve deeper into the cause of this shortcoming, a profound realization has dawned upon me.
Models are often based on the presumption that actors in a market make rational choices. This is, of course, not always true. Fear and paranoia contribute to excessive selling as stock prices drop, kicking off a vicious cycle, for example.
Only the most impeccably hedged quant hedge funds managed to make a profit as the stock market crashed. Others complained that an event that their models predicted would happen once in a millennium happened for four consecutive days.
The complaining managers had no shortage of intellectual firepower or computer whizzes at their disposal. After all, almost all of them had PhDs from the world’s best universities. What they failed to realise was that the one thing that is predictable is that human behaviour is inherently unpredictable!
Humans necessarily abandon reason and logic from time to time. Altruism is a good example. Perhaps our models need to be modified. I am intrigued at the possibility of modelling human behaviour, something so emotional and often illogical using the cold, dispassionate, precise equations of mathematics.
The first challenge is, of course, that given the same situation, each individual may react and make decisions differently. Can we determine behavioural patterns that are statistically significant and model them with the flexibility of expecting black swans to add contrast to the lake otherwise filled with predictable white ones? We must learn to imagine the seemingly impossible.
MAAZ ALVI
Lahore