stocklib
Within stocklib.tcl are procedures to scrape and analyze tweets for stock data.
- Stocks are partially vetted:
- Price Must be at least $10, but not more than $60 per share
- Volume must average at least 250k shares per day
- Individual tweets are evaluated for context and confidence.
- Key phrases are awarded positive and negative points.
- Tweets are locally cached so twitter's limits (API calls, number of results) can be bypassed.
- stocks are ranked relative to one another via lower bound of Wilson score confidence interval (95% confidence).
Results
Statistical analysis of 3 months of data shows that 75% of the time, an high value of 0.5% over opening price is seen. Algorithm improvements are pushing closer to 80% success.
Future Goals
Account for volatility and remove stocks with too low of volatility.