CEU Electronic Theses and Dissertations, 2025
| Author | Pal, Bence |
|---|---|
| Title | Trading Bitcoin With Reddit Sentiment: Strategic Capital Allocation Using ML-Enhanced Technical Signals |
| Summary | This thesis shows that Reddit crowd mood can sharpen short-term Bitcoin trading when us- ing an appropriate model. The sentiment of three-hundred-thousand finance-subreddit posts from 2017-2024 is analyzed and fused with standard price- and volume-derived indicators. Two empirical tracks test the idea: (i) Random Forest classifiers enhanced with sentiment, (ii) a simplified MACD crossover filtered by machine learning gates that use sentiment features. Both tracks are evaluated on the model-level and the strategy-level. I then statistically quantify the added value of sentiment in Track 1, while also examining how it affects the model’s predic- tions. I find that the sentiment-enhanced Random Forest model outperforms its indicator-only counterpart, while the sentiment-based gates’ edge seems to stem only from lowered trading costs. It is also shown that when strong sentiment is paired with high activity levels, we get strong and reliable predictions indicating that Reddit sentiment is a valuable metric in model building. Although the Sharpe-ratio increase becomes statistically insignificant once sampling variance is considered, the evidence still indicates moderate support for sentiment improving forecasting models. |
| Supervisor | Posfai, Marton |
| Department | Undergraduate Studies BA |
| Full text | https://www.etd.ceu.edu/2025/pal_bence.pdf |
Visit the CEU Library.
© 2007-2025, Central European University