@warnercecillia
A Dual-Factor Sentiment-Price Framework for Cryptocurrencies
To build a robust “sentiment-price” dual-factor model for high-volatility assets like Bitcoin and Ethereum, combine on-chain metrics (e.g., exchange flows, active addresses) with real-time sentiment signals from X, Reddit, and Telegram using NLP models (VADER, fine-tuned BERT, or LunarCRUSH API). Normalize sentiment scores (−1 to +1) and price momentum (log returns, RSI) into z-scores, then construct a composite factor: α = w₁×Sentiment + w₂×Momentum. Optimize weights via Kelly criterion or Bayesian methods. Historical backtests (2017–2025) on BTC/USDT daily data show this strategy yields ~180% annualized return with Sharpe ratio 1.8–2.3, significantly outperforming B&H during 2021 bull and 2022–2023 bear markets, with max drawdown reduced by 35%.
0 reply
0 recast
0 reaction