I just added the following disclaimer to my ‘About’ page because I thought by writing this blog I might encourage someone to do something stupid:
“Disclaimer: I know this is a cliche but I had probably better say it: nothing on this blog constitutes investment advice. Furthermore, I would advise 99.9% of people do not even think about building systematic trading strategies for themselves or use online broker system builders. Your Computer Science / Physics / Maths PhD will not protect you from psychological biases.”
I really mean this. The only thing that can immunise you from pernicious behavioural biases are painful losses in the market. It’s called operant conditioning.
Systematic trading is particularly mentally toxic. Armed with a computer and a large dataset, you will be able to find anything you want to see. The market is like an evil magic mirror, it will show you what you want to believe and tempt you in. Using quantitative methods only makes you better at finding whatever it is you want to see. It happens to social scientists all the time. 5% of published research is by definition spurious correlations. But a lot more is ‘tortured data’ with outliers removed, survivorship bias in experimental methods etc, etc.
Now imagine these temptations, but with a huge readily available dataset. Add in the self-confidence of a numerate savant, the smell of money, and some great looking backtesting results and it’s not surprising that some people just lose the plot. I’m not saying that you are not aware of overfitting (bias/variance in ML), and the steps you can take to eliminate it, or the fact that the world simply changes over time making historical data redundant.
What I’m saying is trading, and especially systematic trading can make your limbic system (emotional brain) completely overpower your prefrontal cortex (rational brain) even when you know all of this. You will not believe the delusions your limbic system can conjure up to get what it wants, especially once you step outside the sobering constraints academia.
Now, think about the influx of intellectual and financial capital into ‘big data’. I have no doubt that as in finance, some people will strike gold. I know these guys are trained in avoiding the pitfalls of data mining. But if you build a big enough data set and throw enough computing power at it, you will create an evil magic mirror, especially if you have dreams of wealth and fame riding on it.
Entrepreneurs with technical ‘yes men’ are particularly at risk.
You have been warned.
Aldous Birchall 