Sentinel Alpha detects market narratives as they form, analyzes crowd psychology, predicts moves across three time horizons — and publishes a daily self-grade of every prediction it makes.
Four times each trading day, the engine sweeps social signals, news flow, and market internals — and distills them into discrete pieces of evidence.
Evidence is clustered into the stories actually moving markets. Narratives are ranked, tracked through their lifecycle, and tested against price action.
Every pre-market, the deepest analysis layer asks what the crowd believes, what it's missing, and where positioning has drifted from evidence — then issues predictions on three time horizons.
After each close, a grader scores past predictions against what actually happened — publicly. Systematic misses feed back and recalibrate the engine.
Sentinel Alpha grades its own predictions against actual market outcomes every trading day. Nothing is cherry-picked: the trailing record below is computed from the same grader files the engine publishes.