Finally found structured crypto sentiment data that isn't garbage

I've been building quantitative trading models for about 3 years. The hardest part isn't the model architecture - it's finding clean, structured input data.

Tried scraping Twitter/X sentiment myself. Nightmare. Rate limits, bot accounts polluting the signal, and the NLP models I trained kept confusing sarcasm for bullish sentiment. LunarCrush is decent but expensive at scale. Santiment's API is solid but $50/mo minimum.

Found a pre-processed sentiment feed that scores top crypto assets from 50+ sources with confidence ratings. The data is already normalized - timestamp, asset, sentiment_score, signal direction, confidence level, and even narrative tagging (which theme is driving the sentiment).

I backtested it against 90 days of BTC price action. The sentiment_score divergence from price (sentiment dropping while price holds) predicted 3 out of 4 major pullbacks with 12-24 hour lead time. Not perfect, but that's a 75% hit rate on a $5 dataset.

The data also includes TVL delta and narrative momentum scores, which I'm now feeding into a multi-factor model alongside my order flow data.

If you're building trading bots or quantitative models and need a cheap starting dataset for sentiment analysis, this is probably the best value I've found. The AI agent that generates it apparently reprocesses every 4 hours.

Not affiliated, just sharing because I spent weeks looking for exactly this.

Link: https://hokedev.gumroad.com/l/rhbles