Cashcaval

Team Number: 37 Main file: sentiment_analysis/submission.ipynb Video link: https://drive.google.com/file/d/1eH5bUa7TviNZgYflWdE0zL6RGcvrG3Dh/view?usp=sharing

What it does

The algorithm executes an endless loop, which does the following:

1. The news titles are queried and from them, we check whether a relevant company name is present and assess the sentiment of the title using the BART model.
2. If relevant news exist, we greedily buy (if sentiment positive) or sell (negative) as many stocks as we are allowed using IOC. This stock is then disabled for some seconds (unless it appears in other news), and then it can be used again.
3. Then we check whether the remaining stocks and their dual contain discrepancies which we can exploit. We then buy and sell them with the given prices until the discrepancy disappears or we reached our limits.
4. The remaining stocks are traded normally with the base algorithm.

Challenges we ran into

• Tuning the parameters correctly in order to swiftly buy/sell and have good profits after the pause.
• Executing the code as fast as possible in order to be ahead of the competitors.

Accomplishments that we're proud of

• Managing to have all of the logics run correctly
• Leading the market, when not everyone is active :D
• Writing clean code

What's next for Cashcaval

• Improve sentiment analysis to get a metric for the strength of the sentiment
• Improve the speed overall to manage to buy faster then competitors

Built With

• optibook
• python