I. FOUNDATIONS
The ideas for what would become Muxing Trading started bubbling in our minds around 2012. Having worked in Finance and Trading for quite some time, we had the background and experience to formulate the key concepts for an automated trading system which could monitor risk and manage order flow. Market makers at proprietary trading firms had long provided liquidity to the futures markets by placing orders and allowing for multiple levels of price discovery.
Fundamentally we needed the application to monitor risk and manage order flow. The risk management structure that we envisioned consisted of using principal components analysis to monitor position and the type of risk exposure we had on at any given time. The structural risk was decomposed into different factors such as shift, tilt and the bowing of the futures yield curve. The evolution of the curve relative to our positions determined the trade execution programmatically. These ideas and algorithms would later be constructed into software.
II. TRADING
During the next few years, we joined a proprietary trading firm which allowed us to start a desk and develop these methods with a crew that we hired. At this time, several outfits in Chicago offered these types of arrangements with trading groups who together traded as a conglomerate of desks. We also had an agreement that allowed us to keep the rights of what we developed. All in all, it seemed like a unique and tremendous opportunity for us to venture out and try to build our business there. We employed traders and developers to write the execution code and paid all the desk, trading and data fees necessary for us to operate with the firm. By 2014 we were live and started scaling up operations. Within a year our applications were managing 50M of risk and thousands of trades per day. Luckily, we worked with excellent developers that had the background and experience to functionally handle this volume and stay compliant with the exchanges.
III. FUND MANAGEMENT
Success always piques interest. For us it was a tremendous feeling of accomplishment after working around the clock to get the strategies developed and operational. The firm we were working with decided to open and operate a fund which would have a different business model other than the proprietary trading shop structure. Naturally, many people were worried about what this new structure would mean for their desks and the businesses that they had already started there. We were asked to participate in discussions with an asset management group who was interested in us developing some new fund strategies. It was during this period that they were also probing into the desk strategies and the risk management techniques that we used to manage our portfolio. We did explain the strategies some of our techniques and developed some algos for this new fund entity.
IV. SYSTEMIC CHANGE
Although our strategies did an excellent job at making markets and managing risk, they did rely on market volatility and liquidity to be able to execute orders. During 2015 and 2016 we saw a tremendous change in the futures market and had to evolve our strategies to accommodate them. The one risk we did not factor in was exogenous changes to market conditions. The VIX dropped to an all time low and lot size increased on average during a period of consolidation as smaller market makers left. Price discovery was now set by only some of the largest trading firms in the world. We needed to be able to compete during this time and that meant that we would have to increase our seed lot size and assume more risk. Eventually this dynamic would prove to be unsustainable as we had to increase our positions to stay competitive while simultaneously losing the market edge and execution.
High frequency trading groups were also the target of political campaigns and had long been viewed as evil institutions that serve only the greedy purpose of making money. An unintended consequence of that political warfare was the mass consolidation of algorithms developed by those who had the means and resources to stay solvent. Due to this and the raising of desk and data fees, we decided to stop operating the desk.
V. CONCLUSION
Closing the desk was an incredibly difficult decision. We closed our account with the firm and the remaining funds were returned to us. I am happy that we could sustain a profitable enterprise until the end. We still have the rights to all that we developed, and I have worked on revising the code base to incorporate newer techniques such as machine learning algorithms into our strategies. I redesigned the underlying data processing to handle trading at whatever latency setting whether high or low. The dissolution of the desk allowed me to enter a period of free experimentation and design. We now continue to run the code in simulation, collect data for research, and use it sparingly but who knows what the future holds.
Comments