Hey everyone, I’m Pavel. I’ve dedicated my entire career to the field of algorithmic trading, developing trading strategies and software tools for traders. On my website, I share my insights, ideas, and open-source tools with anyone eager to improve their skills in algorithmic and grey-box trading or just looking to discover something new in this field.

Here is my short story.

My interest began in 2005 with a student project focused on automating decision-making processes in microelectronics technology. Around the same time, during a course on probability theory, I learned about stock market speculation and, more specifically, the Forex market. Shortly afterward, I discovered the MetaTrader 4 platform, which featured integrated algo-trading tools and its own programming language. Importantly, it also had a backtesting engine that allowed for the testing of strategies using historical data.

I began experimenting with simple trading models and, by the time I graduated in 2007, I had developed several strategies with which I traded real money. I even managed to sell a monthly subscription to one of them. During this period, I also published my very first article on trading, which discussed the integration of various indicators – it’s still available here.

I was fortunate enough to land my first job at a US-based algorithmic proprietary trading fund. This was well before it hit the mainstream, but we developed an AI-based generative framework. The idea behind it was that the framework would generate trading algorithms by mixing hundreds of technical indicators, finding the optimal combinations in the process.

Expectations for miracles from AI had to be met, and over approximately a year, we developed a set of remarkably profitable algorithms – achieving around 150% in annual compounded returns. Everything seemed perfect, except that we launched live trading right a few weeks before collapse of Lehman Brothers, which sent global economy tail spinning. Despite AI’s capabilities, the management team, responsible for execution, decided to remove stop losses, banking on the market’s eventual recovery. And as we know – it indeed recovered, but not before we had lost about 40% of our clients’ funds in just one week and everybody including me was fired on the spot.

My next job was with a company that developed software tools for algo-traders, including charting, backtesting, optimization tools, a live trading server, and even proprietary database technology. Due to my previous experience in trading strategy development, I integrated well into the company. I participated in various activities from programming and testing to product management and documentation. It was a great company with great people, and I learned a lot from them. Additionally, for the first time, I was working with customers seeking software tools to either start or enhance their algo-trading business. I also participated in several research projects for company clients; here is an old article I authored that describes one such project.

I particularly enjoyed helping traders solve problems that I had previously racked my brain about, first in my dorm room and later at a hedge fund. I discovered that traders typically attached to one, maybe two, of their own ideas – their unique trading approaches – and apply these to different strategies, instruments, or markets. Their methods could be straightforward or complex, rooted in pure statistics or driven by gut feeling. However, the more they traded, the more refined their approaches became. In other words, the more they learned from their actions, the better traders they evolved into.

After a few years, I was promoted to the position of product manager and received my first project – a grey-box tool which, as you can guess from its name, allowed for the combination of black-box and manual trading approaches.

In my experience, and from what I’ve seen working with different companies, grey-box tools represent the most advantageous combination. In other words, while repeating backtesting and optimization can yield a statistically proven profitable model, if you keep your eyes on every P&L peak, I would argue that the model would become even more profitable and stable.

Not long after, I was offered an opportunity to join the founding team of a new company focused on software for technology-driven traders. We developed software tools designed to tackle the contemporary challenges of electronic trading – they work quickly and easily with all sorts of large historical data sets, support the programming of complex trading models, execute in real-time with microsecond latency, and provide advanced visualization. Additionally, we developed advanced, fully automated risk management software. We chose C++ as the programming language for its computing speed and cross-platform capabilities. Another distinguishing feature was our web-based user interface, which we built using React JS (which was at version 0.10.0 back then).

The harsh truth I was about to learn that in technological business, the technology itself constitutes only about 5% of the entire business. Despite we released the first stable version of our product in about 10 months, we lacked the knowledge on how to sell such advanced software and did not have a marketing budget from our investor, and after about 2.5 years we shut down the company.

Before long, my partners and I launched a new venture based in New York City, focusing on custom development and technology consulting for tech-driven trading firms. We focused on implementing some of our ideas within clients’ trading environments. Our consultancy began with small proprietary shops and trading startups, and eventually, we secured consulting projects with large corporations, such as Neas in Aalborg, Denmark, Uniper in Düsseldorf, Germany, and Northpool BV, in Leiden, Netherlands.

Here are some project highlights:

For Neas’ algo-trading department, we developed a historical data management system, a comprehensive suite of tools for collecting, cleansing, storing, and distributing pricing and weather data tailored for energy trading. This system was integrated with their existing in-house backtesting and statistical tools.

For Uniper, we designed an algorithmic trading framework to enhance their manual energy trading infrastructure. This included developing an API for strategy development, creating a number of execution algorithms, and deploying a simulated trading environment.

For a US-based proprietary trading firm, we conceptualized and implemented a market-making framework for cryptocurrencies. This framework consists of a set of tools for both backtesting and real-time execution, capable of sending hundreds of limit orders per minute

As of now, we offer consultancy services to a diverse range of clients involved in energy trading, stock and FX trading, and cryptocurrency trading.

Now, after nearly two decades in algorithmic trading, I have decided to take my work to the next level. I’ve encountered a variety of interesting challenges and gained diverse experiences across different trading environments — from major firms like Uniper and Northpool to small proprietary funds and individual traders. My dedication has been to the design and development of software tools that I believe can refine, expedite, or enhance research, strategy development, or live trading. I want to share these practical insights through my articles and offer some of my projects as open-source resources for my readers.

Here is my first open source tool: DT-Box. This tool is designed to enhance your manual and grey-box trading skills through visual backtesting. Here I describe my ideas behind this tool, here you can find my article about how this tool can help you grow from an idea to a profitable trading strategy. Then here you can find a guide on how you can install DT-Box and here there is a comprehensive user guide to the application. And finally here you can get the source code, which you can customize to fit your requirements.