What it does
It provides liquidity by quoting both buy and sell side of the traded stocks. It also analyses social media posts using a sentiment analysis model to help us better predict the movement on the market.
How we built it
We started with the basic algorithm provided in the environment. We expanded it by:
- Increasing the throughput of the tool
- Introducing sentiment analysis to our algorithm
- Improving the sentiment analysis speed and accuracy
- Introducing safety measures to prevent holding the positions for too long
- Adding volatility analysis
Challenges we ran into
It was hard to measure our performance and look for bugs. The performance of our code was directly influenced by the strategy of other teams’ bots, so it was hard to determine whether the profit was caused by our improvements or the mistakes of others.
Accomplishments that we're proud of
At some point the frame of our algorithm was really promising but we couldn’t adjust the parameters to make it profitable. We came up with an idea to make a version of algorithm enclosed by a optimisation algorithm tweaking the parameters based on the PnL difference. Eventually we were able to find quite good and reasonable values.
What we learned
We discovered both theoretical and execution parts of market making. Writing a market making bot was a totally new experience to all of us and an amazing learning opportunity. Apart from that, we also had to perform a lot of data analysis. Most of the time in our team was spent trying to find patterns in the provided dataset.
What's next for MM Trading Algo - Team 4
I would love to try out some of the ideas we didn’t have a chance to properly implement and test. I would also read on some stock’s theory and try to utilise this knowledge in the future challenges.
Log in or sign up for Devpost to join the conversation.