Every prediction. Every outcome. Live.
Bloomberg reports. Frontberg predicts. Here's the full track record — machine-measured, unedited, public.
| Timestamp UTC | Causal event | Mechanism | Affected assets | Conviction | Direction | Horizon | Verdict |
|---|---|---|---|---|---|---|---|
| Loading predictions… | |||||||
How we score predictions.
Every interpretation lists confirmers (events / price patterns that validate) and invalidators (counter-evidence that kills the call). Both must be non-empty before AlphaFlow ships a prediction. No call ships without an exit condition.
Verdicts are graded against the prediction's named horizon — immediate, intraday, or multi-session. The grade comes from market data, not the analyst who wrote the call.
A multi-session call is never blended with an intraday call. Each horizon has its own scorecard. Mixing them hides where the engine actually has edge.
When confidence falls below threshold the engine returns nothing instead of a weak chain. That's why the ledger reads tighter than a typical sell-side feed — noise is filtered before it enters the record.
Outcomes ledger lives in engine_resolved_outcomes. Mechanism vocabulary is closed (no free strings). Banned vocabulary ("could", "may impact", "sentiment") is filtered at validation.