Exclusive Interview with Kris Longmore | Founder of Robot Wealth

Kris Longmore is the founder of the online systematic trading community RobotWealth.com whose mission is to provide DIY traders with the support and the knowledge that can be found in a professional trading setting. Kris Longmore left behind a ten-year engineering career to tackle the world of finance and trading. Over the years that followed, he became a hedge fund quant and later consulted to some of Asia-Pacific’s biggest fund managers on the adoption of artificial intelligence and machine learning. He is now a partner and shareholder of a proprietary trading firm along with founder of Robot Wealth. Longmore focuses on searching for new sources of alpha using the most exciting and cutting edge tools of the modern age, like machine learning and artificial intelligence.

Kurt: Could you just tell us a little bit about your background?

Kris: My background is in mechanical and environmental engineering. I spent a decade working in that field before I got into finance, doing things like computer simulation of environmental processes. I was into the design of water-related infrastructure as well.  In the finance world, I’ve traded for hedge funds and I founded a quantitative analytics consulting company, and now I trade and research full time as a partner for a prop trading firm. I also founded the online systematic trading community RobotWealth.com whose mission is to provide DIY traders with the support and the knowledge that can be found in a professional trading setting.

Kurt: You mentioned that you come from an engineering background. How did you first get into trading and what was that experience like?

Kris: I was introduced to trading by a friend of mine who had been trading for a little while, who showed me a technical trading system, and I was going through a process of implementing it manually. At that time, I was a complete market newbie and I was really intrigued by the prospect of windfall profits but I couldn’t really come to terms with the lack of evidence surrounding the strategy and whether it actually worked. Over the course of a couple of weekends I painstakingly back tested it by hand and I had to conclude that it was bogus. But during that process I got totally hooked on this systematic evidence-based approach to essentially making money in the markets. And that really began a multi-year journey of learning and experimentation that ultimately led to some private investment backing, trading, and taking a development job at a hedge fund.

It was really difficult obviously but it was also quite rewarding. The nature of what we do with trading systems research is that you really need to be comfortable with rejection as most of the strategies and ideas you come up with will just not pass. You become attached or biased to a strategy because you’ve invested a lot of time on and that can wind up being an expensive exercise.

You become attached or biased to a strategy because you’ve invested a lot of time on and that can wind up being an expensive exercise

Kurt: When you were making that transition from engineer to trader, what were some of the hurdles that you faced?

Kris: I guess the biggest one was probably the lack of time. At the time I was also running my own small engineering consulting company and so I didn’t have a lot of spare time. But what I had, I devoted to this trading hobby that I had picked up. And I was really lucky to have a supportive partner who is now my wife, who really allowed me to spend the time I needed to on it and I couldn’t have done it without her, so I was lucky in that regard. I also tended to place too much confidence in what other people said about the market and what works in trading systems. I had this ‘eureka’ moment when I stopped accepting what other people claimed, no matter how innocuous it might seem, and I started exploring those assumptions with my own research. That was a real turning point.

The lack of mentoring was a big obstacle as well. I wish I could have bounced some of my early ideas and got some practical advice from people who had experience in the market. When I finally started to make some consistent trading profits I assumed that what I built was nothing really exceptional and that it was quite normal in professional trading. I showed my results to someone I met who happened to have some connections in the finance world, which was at that time completely new and foreign to me. Before I knew it, he connected me to a high net worth individual who wanted to back my system. I was obviously really surprised about this but I went ahead and did that for a while and that led me to this offer to join a hedge fund. I remember how surreal it felt to be kind of courted by this industry that I’d totally put on a pedestal.

Kurt: Over time, I’ve seen your website Robot Wealth evolve. Could you tell us what inspired you to start this website and some of the benefits traders can get from it?

Kris: Sure, it’s really interesting because when I first started blogging on Robot Wealth it wasn’t a community at all– it was a place where I was documenting part of my research. I was really trying to tap into some of the knowledge out there so I started posting my research to get feedback. It really went from there. There was some interesting stuff that the audience appreciated and so I started documenting and then writing a lot of the things that I learned during my career in professional finance and then I ended up writing a couple of algorithmic trading courses. The purpose of those was to put in one place all of the things I wish I knew when I first started. And so I started offering those for sale on the website and people got really interested in that. Then it seemed like a natural progression to provide an online place to build a community and collaborate with each other and share the journey together. What is really interesting about Robot Wealth and the retail trading space is that retail traders typically don’t have access to the support, tools, and knowledge that you find in a professional trading house. Professional traders get access to a wide set of skills through a team, and really useful but expensive data.

And so you have this gap between the environment that retail traders find themselves trading in and that of the professional traders. The goal of Robot Wealth really evolved into becoming a place that levels the playing field by providing those same tools and support that you find in your professional trading house.

The goal of Robot Wealth really evolved into becoming a place that levels the playing field by providing those same tools and support that you find in your professional trading house.

Kurt: What kind of barriers of entry do traders need to overcome to get involved in algorithmic trading and ultimately implementing machine learning in their trading?

Kris: Well I guess programming skills are the first and the most obvious prerequisite. I learned that the hard way when I manually backtested, but then I realized there must be a more efficient way and there is. It is easy to underestimate just how much work goes into an algorithmic trading system. Just writing an accurate backtester is incredibly difficult and time consuming. You need to worry about linking your trading system to an execution venue, automating as much of the monitoring and reconciliation process that you can, not to mention all the research and testing that goes into developing the system in the first place. I think the barriers to entry are number one programming skills and number two finding the time and the space to do all of that.

Kurt:  To use machine learning and artificial intelligence, what kind of a workstation do you recommend for traders?

Kris: We don’t really need an exceptional set up to get started with machine learning. For example, I train some of my deep learning models on a laptop. These days when you run out of computer resources it is quite trivial and relatively inexpensive to get all of the computer power that you need using services like AWS or Google Cloud. I use a variety of tools for machine learning and AI. For performing machine learning model tuning, I really like a package you can find on the R program called caret. It is one of the one of the best machine learning packages out there, I think.

Kurt: What we’re seeing is that artificial intelligence and machine learning are becoming more adopted on both short and long time frame. What effects do you think this can have a market looking forward?

Kris: That’s an interesting question. Machine learning and artificial intelligence really enable us to incorporate a large and diverse array of information in our decision making process. To the extent that these models are able to accurately determine the fair value of the market is only going to make the market more of efficient over time. But having said that, machine learning and AI simply are the latest evolution in the search for alpha extraction and information advantages. This stuff has been going on as long as markets have been in existence and we’ll certainly see some sort of alpha drying up as the result of these technologies, but other opportunities will arise from consequences as well.

Kurt: What effect do you think quantum computing will have on machine learning and finance over all?

This is a super interesting area. First and foremost, quantum computing is going to enable the processing of more data at a faster rate. If you think about it data is really the fuel that drives machine learning and artificial intelligence. Following that logic, it leads us to believe that quantum computing is going to provide a step change in the performance of machine learning and AI.  Another use for quantum computing is going to be the fast, online solution of complex optimization problems and that has some very obvious and important implications for portfolio management and some pricing models. I really would not be surprised if we find ourselves with markets that start to really push the envelope of the efficient market hypothesis. At this point and when we get there, it’s going to be quite difficult to compete without using such technology. But thankfully it’s a few years away.

Sometimes if you reframe the question a few different ways you can provoke some inspiration.

Grace:  What are some of the best practices for feature engineering?

Kris: With the exception of deep learning algorithms, the feature engineering process is in my opinion even more important than the actual model tuning process. If you have engineered good features, you will generally get good results in machine learning algorithms regardless of the choices you make with respect to the model selection and tuning. Such decisions are still important but they tend to deliver incremental improvements compared with good feature engineering. The whole point of feature engineering is that you transform your input data into something that the learning algorithm is going to understand within the context of the problem you are trying to solve. Feature engineering is really a huge topic, but some high level tips that I can pass on are things like spend some time thinking quite deeply about the problem that you’re trying to solve and brainstorm features that might be predictive. Sometimes if you reframe the question a few different ways you can provoke some inspiration. If you can find them, talk to experts in the area of the problem you are trying to solve as well. They can often be a really great source of inspiration for input and modelling.

Grace: In regards to techniques for avoiding overfitting what are some that you would recommend?

Kris: It really depends on the application but some approaches include things like avoiding brute force optimization where possible, and can include regularization techniques which essentially is a mathematical trick for shrinking the coefficient associated with some of the variables. Also penalizing complexity as well in a modelling task. It’s a tricky one but sometimes just identifying overfitting can be useful and the usual tools like walk forward analysis or time series cross validation can at least help to identify when it’s occurred.

Grace: Along similar lines, what techniques do you recommend using to avoid data mining bias?

Kris: In practical terms, it’s really unavoidable when it comes to research. It’s an issue in traditional research processes where we might manually inspect the results of the same out of sample set a few times but it becomes a really massive issue and we start to use these modern computational tools to mine a dataset many times over. How to deal with that really depends on the application, but if it’s possible you can use what I call a true out of sample set once and once only to get an unbiased estimate of performance. However that’s not really an efficient use of data and of course our market data is finite so really not a practical solution. In some applications you can also use include statistical resampling techniques to quantify the impact of data mining bias.

There are some authors who recommend deflating the performance metrics from a backtest based on the number of different strategy variants you looked at. The approach is a little bit crude but it’s an effective and simple way to deal with the data mining. In some applications, it can also pay to take a different perspective and embrace data mining bias. You can do that, for example, by analysing every possible strategy variance in order to understand the distribution of past possible performances.  That’s not going to be applicable in every instance but it can work with some strategies. And none of these approaches that I’ve mentioned are really perfect by any stretch but are really useful from the perspective of compiling evidence for taking an idea to market or not.

….to summarize that advice I would say be ambitious but patient.

Kurt: Are there any books that you recommend for traders?

Kris: I’ll go out on a bit of a limb with this one and it’s one I can’t speak highly enough of for trading, it’s called Thinking Fast and Slow and it’s written by a psychology researcher by the name of Daniel Kahneman who I believe has won a Nobel Prize.  It’s not about trading specifically, although the author has an affiliation with the field of behavioural finance. It’s a book about understanding the way that the human brain works and how that influences our decisions and biases through which we see the world. I think it’s a great book for anyone interested in tackling the markets.

Kurt: Before we let you go, is there any advice you would like to share with traders that you wish you had known prior to embarking on this long journey?

Kris: There is quite a bit I wish I would have known, but if I were to summarize that advice I would say be ambitious but patient. Be highly critical of assumptions and claims in real life, and first and foremost do your own research. Seek out sensible shortcuts but recognize that there isn’t a shortcut to everything– you do have to do a lot of hard work. Try to pick the brain of people who’ve gone before you and really proven themselves. Embrace technology and most importantly never stop learning.

Get Started Algo Trading!

Create your own API-enabled demo account and start your journey today. To generate your free API token:

  1. Register for a free practice account here.
  2. Log into fxcm.com and click on the account ID in the upper right corner.
  3. Click token management and generate your token.

Risk Warning: The FXCM Group does not guarantee accuracy and will not accept liability for any loss or damage which arise directly or indirectly from use of or reliance on information contained within the webinars. The FXCM Group may provide general commentary which is not intended as investment advice and must not be construed as such. FX/CFD trading carries a risk of losses in excess of your deposited funds and may not be suitable for all investors. Please ensure that you fully understand the risks involved.