Deterministic observations over live market and trade data. Research rows track movement patterns; resolution rows track settlement risk. Every row stays a verification workflow, not a trade call.
Total firing
45
Momentum
9
Divergence
2
Flow
0
Resolution Risk
34
Aggregate after-spread verdict per signal kind, mirroring the /backtest page. The inspector on the Research tab also shows the live slice verdict per signal so users can see when the live observation carries higher evidence than the historical aggregate would support.
Weak after spread on /backtest: ~-2.6pp expected 24h move, ~40% win rate across the top-15 sample. Treat as research, not a profitable signal.
Live measurement (748 firings, momentum): mean 24h drift -10.8pp · mean |drift| 18.3pp → STRONG.
Weak after spread on /backtest: ~-1.8pp expected 24h move, ~49% win rate across the small sample available. Inspect, not actionable.
Live measurement (306 firings, divergence): mean 24h drift -11.1pp · mean |drift| 18.4pp → STRONG.
Forward-only — flow signals need persistent trade-feed history we don't yet store.
Forward-only — news ingestion isn't persisted yet, so there's no backtest to aggregate.
Forward-only — event-time risks can't be replayed from price history alone.
Live measurement (874 firings, resolution_risk): mean 24h drift +1.4pp · mean |drift| 17.3pp → MIXED.
Verdicts are a static snapshot from the latest /backtest run on the top-15 markets by volume; they drift as upstream data changes. A persistent backtest history will replace this snapshot once the signal-firing ledger lands.