Contact:    Log in
Raw, unmitigated computer power at your service
Raw, unmitigated computer power at your service
How Quants failed Wall Street but is redemption in sight with the arrival of AI?
Published on 31 May 2023
Allan Lane
Allan Lane
Algo-Chain, Co-Founder
One of the surprises of working in finance is how little innovation there is in the world of investing. This is about to change. If you don’t believe me, check out the views of Mahmood Noorani, the Co-Founder of Quant Insight, on a recent LinkedIn post. Working in conjunction with a team of machine learning leaders from Cambridge University, they have created a way to construct a portfolio of stocks (50+ typically with no stocks more than 5% weight) that track other variables of interest. For example, they are able to track the CDX HY Credit spread with an 87% daily out-of-sample accuracy. To achieve this, they created a Bayesian optimizer that solves 10^100 combinatorial problems in less than 1 minute!

Nothing beats a bit of hyperbole to get some attention, but I think Mahmood is on to something here. Taking a high-level overview of the last 30 years of how innovation has rippled through the investment industry, it does already feel that 2023 will be remembered the year when many ground breaking ideas were turned into reality.

It's barely the end of May, and at Algo-Chain we ourselves have already had to review the stack of 3rd party AI tools that we use to deliver our solutions. New packages are coming to the market daily, often leap frogging the heights achieved by the current status quo. While nearly all the attention has been Chat GPT focused, it is the intelligent use of extreme levels of computer power that is really in play here.

The formative years when quants were just quants

During the early part of the 1990s I was working for Paribas, who, along with the likes of JP Morgan and Société Générale, loaded up on mathematicians and engineers trying to solve the problem of how to build trading and risk management systems that could keep up with the explosive growth of new products coming to the market. Innovation was rife and a team of smart programmers had figured out how to design a new in-house programming language that was able to price and risk manage a wide range of structured note pay-outs all from a single code base. There was no stopping the sales desk, almost overnight the team were able to make prices on new structured products which kept them one step ahead of the competition.
Every trading desk had a quant team playing the role of the Central Processing Unit
Every trading desk had a quant team playing the role of the Central Processing Unit
There have been several occasions when a new idea landed on Wall Street and life as we knew it was never the same again. Myron Scholes along with several other academics brought option pricing to the main stream, and iteration by iteration this gave birth to the Structured Notes industry, which in turn gave rise to the cult of Credit Default Swaps and then Collateralized Debt Obligations, CDOs.

There are wildly varying estimates of how much was lost during the Global Financial Crisis of 2007-2009, where CDOs played a leading role. However, even taking some of the lower estimates, which is in the $trillions, that is some legacy and one that put paid to the notion that all financial innovation is good.

Social networks replace math skills as the weapons of choice

Although technology played a central role in the growth of the structured notes market, what followed was something altogether different. Between the arrival of Bitcoin in 2009 and the ubiquitous role that Mobile technology came to represent, one had the ingredients for a perfect investment bubble.

Unlike the bubble pre-2008, this one wasn’t built on quant innovation, instead take three ounces of a cryptography algorithm and 100 million Bitcoin influencers and before long exponential returns are rolling off the digital ticker tape. It’s reasonable to suggest that without the power of social networking, crypto currencies would never have become so successful.

Behind the scenes though, all was not well. The concept of leverage was borrowed from a decade earlier, but this time round some investors seemed happy to invest in leveraged Bitcoin products touting returns of 10x or higher. This combined with the off the scale level of dishonesty that FTX (a crypto exchange & hedge fund) came to be known for, and before long there was ‘trouble’ at the mill.

Those new investment ideas had once again not shown the investment industry at its innovative best. How else can one explain the comments from FTX’s current CEO John J. Ray III, who said "Never in my career have I seen such a complete failure of corporate controls and such a complete absence of trustworthy financial information as occurred here."

Pure computing power rules & the onset of the golden age of quantitative innovation

In a world where immense computing power is on tap (yes, I’m referring to the Bayesian Optimizer that Mahmood Noorani is putting to good use) then anything is possible. Most portfolio problems are combinatorial in nature, and making sense of how those outcomes behaved in the past and might behave in the future, is one of the key benefits that Machine Learning Models can offer.

Given access to what feels like unlimited computing power, does suggest that very soon we will have the ability to synthesize the returns of any realistic market variable by trading a set of assets, such as single stocks. In which case financial product development teams will have a field day as financial contracts will be written that can offer much more predictable returns. What would you rather invest in, a product that most of the time pays 8% per year or one that sometimes pays 4% and sometimes pays 12%?
Tomorrow's alpha is all about the ability to control the potential that AI offers
Tomorrow's alpha is all about the ability to control the potential that AI offers
With the intelligent application of raw computing power, often sold under the label of AI, we might well be talking about the onset of a new field of finance that is akin to particle physics. In this world real objects are manifestations of 1000s of particles known as Quarks, Leptons and Bosons, which are glued together in mysterious ways.

Already most of us have come to accept we don’t really stand a chance when it comes to figuring out how the large natural language model of Chat GPT is working, but who cares. And so it will be as the benefits of computing power give rise to the golden age of quantitative innovation. Let’s start with the building bricks that comprise the S&P 500 and the Russell 2000, and throw in some commodity futures for good measure, and see what interesting payoffs we can imagine using these basic investment particles.

Until next time.

Allan Lane