Wake up, work, sleep, wake up, work etc. If you’ve ever felt caught in this loop, something must change. Exactly twelve years ago, I quit my last “proper” job at a large firm, Nomura. I really enjoyed what I did there and also before that, at Lehman Brothers. Despite that, I felt that after eight years working of banking, something different beckoned, so I went independent. After each year that passes, I try to look back, to see what I’ve learnt from the experience. Some of the lessons can be the same as previous years, but there’s always something new that I’ve learnt, and a the continued realisation that there’s an ever increasing amount left to learn. I also recently recently visited an exhibition of the Syrian artist, Marwan, who painted the above portraits, which put me in a somewhat more reflective mood regarding the past.
Have a plan, and some balance
I say this, because when I quit, my plan was (nearly) non-existent. My “plan” was to basically continue what I was doing in the bank, publishing research on FX markets, albeit as an independent, at Thalesians, a quant think tank which Paul Bilokon, Matthew Dixon and I cofounded just after Lehman Brothers. I still help to organise quant meetup events at Thalesians with Paul, and I recommend you come to our events, and I’m very proud that the events have been going on for over fifteen years.
What I had negated to do as part of my plan, was to understand whether people would actually buy the research. You cannot simply make assumptions. Ironically, sometimes when you are more quantitatively inclined, it can be tempting to concentrate on your core area (in my case quant research of financial markets), and put all your energy into that. Somehow you can neglect any other areas, which are important when building a business, where a quant approach could be useful, like a systematic approach to sales and marketing. Building a successful business obviously needs a great product, and that is not the only prerequisite.
Pivoting
Suffice to say my original idea of publishing quant FX research didn’t work out. The pre-MiFID world wasn’t a great environment for independent research. I also didn’t market the idea sufficiently. In the end, I lived off intraday systematic FX trading on my personal account. It wasn’t my initial goal to live off my trading income (and I certainly don’t recommend this for most people). The flipside is that it did allow me to “seed” my business and sustain myself. You also learn a lot from trading about markets, in particular the emotional aspect of managing cash. There’s also a different dimension when it’s your personal account, that you are managing.
In the end though, I realised that I needed to pivot, and I started to do consulting under a new company Cuemacro. It was fun to work on many different projects over the years, but ultimately, there’s only so much you can do working by yourself, and felt it was necessary for a change. Hence, I cofounded Turnleaf Analytics with Alexander Denev, my fellow co-author of The Book of Alternative Data. At Turnleaf Analytics we have also done some pivoting, for example where once there was a focus on forecasting of inflation purely for EM right at the beginning, we quickly started doing DM as well very close to when we started the firm.
Your team is the most important part of your business, in particular in the age of AI
Everywhere, we hear about AI, albeit centred around LLMs. LLMs are no doubt very powerful. We regularly use them to understand the market and to help us code at Turnleaf Analytics. We also use AI to forecast time series, although, I would caution that the types of machine learning techniques appropriate for this are somewhat different to LLMs. I think there can be a fallacy that AI alone can somehow solve everything, rather than using it as a way to accelerate your research process.
If we take the example of coding, AI is a great tool for generating code. However, we need to guide these tools. For anyone who has coded they know that the most time consuming part of it is maintaining the code, rather than the initial stage of writing code. If we architecture a system properly, we can make code easier to maintain. Even if we use LLMs, we need to understand the code they generate. Having an knowledge of the code base is important to debug and maintain it.
In this age of AI, the team is even more important than before. If the team can use AI effectively, they can accelerate the time it takes to develop models and solutions. By contrast a team, which uses AI ineffectively, can store up problems for the future, whether that is in maintaining a code base or other areas.
Keep learning and researching financial markets
I remember until a few years ago, listening to Don’t Look Back in Anger, I inexplicably used to assume Noel Gallagher sang the words “So Salican Way”. I’m not sure precisely how I formed that observations. Eventually, I realised the actual lyrics were “So Sally can wait”, somewhat different. There are so many things that our team has learned at Turnleaf Analytics over the years, which was so obvious in hindsight (more about forecasting inflation, as opposed to Oasis lyrics), but seemed totally obscure beforehand.
Every year that has passed at Turnleaf Analytics, we’ve built up an every increasing insight into ways of improving our inflation forecasting. I’m sure that in the years ahead, we’ll discover many more things about forecasting time series which we didn’t know before. Continual research and learning is an imperative part of financial markets. You will never solve the market (when someone says they have close to 100% accuracy, I would question what precisely they are measuring), but you can always try to improve your accuracy.
The human element
As a quant it can be tempting to think that everything can be made into a model, but ultimately a model is some simplified description of the market. It takes some real creativity to think of ways to improve a model and interpreting the live performance. You want to understand for example which data is missing in a model (yes, some data will always be missing, but we can try to reduce the “missing” area). Is there some sort of factor you’re missing? Is there something that our data won’t capture, so we need to estimate it in a different way?
Conclusion
The above observations are not exhaustive. They are just a summary of some of things that come to my mind as lessons I’ve learnt after the past 12 years of independence. What will I learn in the coming year? It is difficult to answer this, although, I suspect it will be obvious once I’ve learnt it. All I do know is that I find financial markets just as exciting as I did when I was in banking!