{"as_of":"2026-09-06T04:43:04.035537+00:00","window_days":7,"overall":48.0,"overall_label":"Neutral","categories":{"crypto":{"score":41.2,"label":"Negative","tickers_with_data":5,"tickers_total":5,"headlines":519},"fx":{"score":48.4,"label":"Neutral","tickers_with_data":7,"tickers_total":7,"headlines":126},"equity":{"score":53.0,"label":"Neutral","tickers_with_data":20,"tickers_total":20,"headlines":3784},"etf":{"score":42.3,"label":"Negative","tickers_with_data":5,"tickers_total":6,"headlines":93},"commodity":{"score":54.9,"label":"Neutral","tickers_with_data":4,"tickers_total":4,"headlines":696}},"methodology":"Built from real financial news headlines scored by FinBERT (a financial-domain NLP model) -- not trader positioning, order flow, or social-media chatter. Score per ticker = mean sentiment over the trailing 7 days. Category score = equal-weighted mean across that category's tickers (headline volume doesn't skew it). Overall = equal-weighted mean across the 5 asset classes, so equities' 20 tickers can't dominate a number meant to represent all of them. Scale: 0 = extreme negative, 50 = neutral, 100 = extreme positive."}