Inspiration
As five physicists from UCL with a passion for complex problems, the Optiver challenge proved to be an amazing opportunity to refine out technical skills and learn more about the world of algo-trading and market-making, while working on the solution of the dual listing problem.
We immediately started our research of possible techniques for market-making, learning the basics from scratch, and attempting a first rough algorithm (which immediately proved very, very, wrong).
What it does
After several iterations, we built an algorithm that continuously holds both a bid and an ask in the illiquid market, making sure that the bid-ask spread for Philips_B that we set is always wider than the bid-ask spread for Philips_A, so that we can always immediately hedge our position on the more liquid market while making a profit. In the long run, as the market randomly moves, the difference in times our bid and our spread are met will tend to zero, essentially cashing in the profit from the initial position.
Challenges we ran into
The main challenge we faced was seeing a model working theoretically , but losing all hopes as soon as it faces the real-world (aka the simulated environment within optibook). To overcome it, we had to reiterate on the model numerous times, following a trial and error approach until finding the most suitable strategy that actually works also in the real-world.
Accomplishments that we're proud of
We are proud of being able to learn so much about a field initially extraneous to us so quickly, and to be able to effectively create a model that could stand a chance in such a complex (and frankly messy) market.
What's next for MAD-LAD
As our coding skills were able to carry us through the task, we realised that we mainly lacked the theoretical knowledge to understand what to code. The next steps will therefore be to really go in depth into the study of algo-trading strategies, understanding better the different complex elements of such a market, and adapting MAD-LAD accordingly.
Log in or sign up for Devpost to join the conversation.