BTC Signals /signals

Loading...

⭐ SIGNALS 0

No signals firing.

Live Prediction Numbers 0 horizons

Waiting for live model output...
HorizonTypeDirectionP Up Raw MoveCalibrated MoveSmoothed MoveTarget ConsensusForward Status
No live predictions yet.
Raw = direct model return. Calibrated = clipped model return used for the official prediction. Smoothed = page EMA so dashboard smoothing is visible instead of hidden.

Live Consensus NEUTRAL

Waiting for signal bundle...
SourceSideHorizonConfidenceSummary
No live signal bundle yet.

Forward-Test Ledger 0 evaluated

No live predictions evaluated yet.
StatusHorizonPredictedActualStartDue
No stored forward-test rows yet.
Every live model row is stored when it fires and scored after its horizon expires.

🔥 KILLER INSIGHTS auto-derived BTC

Computing...
Top edges auto-derived from the latest benchmark’s time-of-day & day-of-week buckets. Updated every benchmark run.

🔌 SIGNALS API for external apps

REST
GET /api/v1/signals/BTC
Returns a single JSON bundle: live market snapshot, AI ensemble per-horizon predictions (1m…8h), all firing god-signals, and a unified signals[] array with plain-English summary + side + confidence for every entry. Cached 10s.
Try it →  ·  Read docs →
WebSocket
ws://<host>
>> { "type":"subscribe", "exchange":"signals", "symbol":"BTC" }
<< { "type":"signals", "symbol":"BTC", "data":{ ... } }
Subscribe and you’ll get an immediate snapshot, then a full push every 15s while subscribed. Same shape as the REST response.
Sample signal
Loading live sample...
Glossary
  • side — BUY or SELL. What to do.
  • horizon — how long until you should be out of the trade (e.g. 4h).
  • win_rate — how often this rule has been correct historically.
  • edge_pct — win-rate above baseline. NOT a price target.
  • price_target — AI ensemble’s projected price at the horizon (only on ai_ensemble signals).
  • confidence — high / medium / low derived from edge size, sample size and p-value.
Internal use. No auth. Plain JSON. Same payload powers the dashboard and the live-evaluator.

Research Reference benchmark

Historical benchmark artifacts live below. They explain why a signal exists; the live prediction and forward-test panels above show what is happening now.
--
Weighted Ensemble · Balanced Acc
--
Member configurations
  • Waiting for /ai/signals...

Research Weighted Consensus per Horizon

Loading...
Variance-weighted mean balanced accuracy across all configs that cleared 60%, with diversity bonus.

God-Signal Sentiment Rules 0

HorizonDirectionThreshold NUp RateEdgep
Run benchmark with --include-thresholds.
⭐ = god-signal (N≥50, |edge|≥10%, p<0.01)

God-Signal CVD Rules 0

HorizonDirectionThreshold NUp RateEdgep
Run benchmark with --include-thresholds.
CVD percentile bands (top/bottom X% of last 168h).

Combined Sentiment + CVD God-Signals 0

HorizonDirectionRule NUp RateEdgep
Run benchmark with --include-thresholds.
Both sentiment AND CVD must agree. Fewer triggers, higher conviction.

Hour-of-Day P(up) Heatmap (UTC)

No data yet.
Green = above base rate. Hover for sample count.

Day-of-Week P(up) Heatmap

No data yet.
Mon–Sun by horizon. Same color scale as hour heatmap.

Lead-Lag Correlation

MetricLag (m)HorizonNCorr
No data yet.
Top correlations between past metric value and future return.

Per-Market Catalog 0

ExchangeSymbolTypeRows
No catalog.
Top markets by row count over the benchmark window.

Working Laws

  1. Loading...