AI sentiment-price correlation experiments for Stocks markets.
## Overview This experiment identifies which news sources best predict Solana (SOL) price movements using a 180-day window with predictive-only article filter. ## Methodology Source predictability analysis with min_quality=0.6 and predictive-only filter (future-oriented articles only). 40 sources analyzed. ## Key Findings The Daily Hodl leads SOL price by +10h with r=0.575 (Strong). FXStreet leads by +4h with r=0.394 (Moderate). Business Insider shows negative correlation (-0.773) suggesting contrarian signal at +24h lead. ## Limitations Correlations based on 180-day window. Predictive-only filter reduces dataset size but improves signal quality.
## Overview Source predictability analysis for Optimism (OP) over 90 days, revealing exceptionally strong oracle sources. ## Methodology Standard source predictability with min_quality=0.6. Only 6 sources had sufficient articles, suggesting OP is covered by a concentrated media set. ## Key Findings CoinGape and AMBCrypto show near-perfect correlation (r≈1.000) at +23-24h lead time, acting as strong price oracles. Crypto news also shows r=1.000 with +24h lead. ## Limitations Very small source pool (6 sources) limits statistical robustness. Perfect r=1.000 values suggest high volatility in the correlation window.
## Overview Source predictability analysis for Ethereum (ETH) over 90 days, identifying oracle and echo chamber sources. ## Methodology Standard source predictability with min_quality=0.6. 26 sources analyzed over 90 days. ## Key Findings PYMNTS leads ETH price by +23h with r=0.624 (Excellent), publishing payment/fintech coverage that anticipates ETH moves. CryptoPotato identified as echo chamber — follows price with -22h lag, r=0.528. ## Limitations 90-day window captures one market cycle. Results may vary across bull/bear regimes.