Inspiration
We are inspired by all the genius other teams that gave us anxiety throughout the whole challenge due to their enormously high score!
What it does
Simple, low/no-risk market-making, sentiment analysis and good old arbitrage :)
How we built it
Using a classifier model to check the sentiment, and then using highly performant business logic to take care of the market making
Challenges we ran into
To make sure the trading bot went smoothly, robust programming was a must. Every edge case had to be considered!
Accomplishments that we're proud of
We've set up an efficient, multi-threaded, notification-based architecture which can outperform many other simpler, polling-based architectures. Our sentiment analysis model has been tuned to be incredibly fast (~0.1 seconds) while having a very low false negative rate.
What we learned
Strategy alone is not everything! Good engineering is just as important as good ideas
What's next for CHANge Optiver
sleep
Built With
- finbert
- python
Log in or sign up for Devpost to join the conversation.