Cameron's Blog - The Rise of DIY Quants
The Rise of DIY Quants      May 16, 2017

Contents

This post first appeared in the May 2017 edition of the Reading University Investment Society newspaper - you can find a copy of the whole paper here.

Modern finance is a constantly evolving field. In the 1970s after Black-Scholes published their seminal paper, derivatives in their current form (for primitive derivatives are found as far back as in ancient Sumerian culture) became common and ubiquitous. Approximately ten years later, we saw the rise of the first quantitative powerhouse Renaissance Technologies, a firm that to this day makes absurd profits. Even today, the rise of passive investing is driven by mathematics as firms construct smart-beta models and strive for mean-variance portfolio optimization.

But the past five years have seen another interesting development - the rise of DIY quants. Idle software engineers, physicists, students, hedge fund managers, and even yours truly have engaged in the construction of systematic and automated trading tools with free tools like Quantopian, Quandl, CloudQuant, Numerai, and many others. These platforms combine free data with programming and software development tools, and many offer tutorials on quantitative investment strategies and techniques.

Take Quantopian for example. In about ten minutes from creating an account, you can have a algorithmized strategy for mean-reversion up and running. You can even hook your algorithm up to two brokers for live trading, either Interactive Brokers or Robinhood. Quantopian offers minute-level equities data, as well as futures prices to trade on, all callable by an easy-to-manage API. Participants on the platform can place their algorithms in competitions, with the prize being capital awarded from Quantopian and its full deployment in the market.

You can couple your Quantopian strategy with Quandl’s vast array of core and alternative data, such as satellite imagery, oil tank storage levels, retailer email receipt data, and other custom datasets. One could imagine a complex algorithm that locates all oil tankers currently shipping, weights the levels in reserve, and prices equities for oil and shipping as well as oil derivatives. In fact, it’s likely that some hedge fund somewhere in the world is already doing such a thing.

The world is becoming more accepting of this type of behavior. In some ways, it reflects global culture’s growing aversion to high finance, as we shift away from large banks and financial institutions and move to passive investing. In the same way, DIY quants are democratizing previously unavailable services and reclaiming sovereignty over their investments.

This type of investing wouldn’t have been able to occur even in the 2000s - few had the skills necessary to write the code, the computers were too slow, the digital infrastructure was lacking, and the data was prohibitively expensive. Python and other high-level languages (often condemned by institutional quantitative investors as being too slow for production) have risen in ubiquity. Even in the late 2000’s the difference in speed between Python and a faster language, like C or Fortran, was nearly insurmountable due to the slowness of the computers they ran on. Now, a budding math-geek can run a trading bot on a cloud server from a Chromebook, and utilize thousands of times more computational power than was used to send men to the moon. Lastly, the data is too expensive! In the past, someone who wanted to do what someone in their garage today could do for free would have to pay thousands of dollars to have CDs shipped to their house in order to model the data - now, all you need is a quick API call and you can have world-class data.

I will note that this type of investing cannot compete with true high-frequency trading outfits. Such firms pay millions to co-locate inside broker’s facilities to reduce latency, and often have the ability of paying near-zero explicit transaction costs. DIY quantitative investing should work on minute scale or longer. I personally have a portfolio management algorithm that simple performs mean-variance optimization and rebalances once a month, so I don’t have to think about my investments.

If you think you’d like to be a cog in the efficient market, go check out any of the quantitative investment sites! Try and find a signal that nobody else has found. If you do, give me a call – maybe I’ll invest.