SignalScope monitors public ticker mentions across eight signal sources — from social media and SEC filings to congressional trades and Polymarket prediction markets — aggregates them by symbol, scores each candidate with AI, runs a 13-flag pump-and-dump filter, and validates signal quality against a LightGBM ML backtesting pipeline trained on historical breakout outcomes across 308 engineered features. The result is a prioritised watchlist of tickers with the strongest multi-source backing, verifiable catalysts, and machine-learning-confirmed signal patterns — surfaced before the crowd.
SignalScope shows two independent 0–100 metrics. Opportunity (early-mover score) ranks how favorable the setup is for timing — validated in ML and used to sort the Signal Dashboard. Signal confidence (AI score) measures how strong the evidence is across sources and sentiment. A very high AI score often coincides with broad agreement; by then, more of the move may already be in the price, so higher confidence does not always mean higher forward returns. Use Opportunity when you care about being early; use confidence when you care about how well-supported the thesis is.
Monitors 17 investing subreddits for posts and high-engagement comments.
Posts + comments · 17 subreddits · 1.5 s delay between requests
Keyword search for ticker mentions from the past 24 hours, run once daily before market open.
X API v2 · 24 h lookback · up to 300 tweets/run
C-suite open-market purchases of $50 K or more from OpenInsider and EDGAR.
C-suite only · $50 K+ purchases · open market only
Flags symbols whose volume is ≥2× their 10-day average.
89 symbols · ≥2× 10-day avg
Detects unusual call volume, heavy OTM call activity, call sweeps, and net premium flow (call vs put dollar volume) across a watchlist of liquid stocks.
89 symbols · Vol/OI ≥3× · OTM 10%+ · nearest expiry · net premium & call/put ratio
Trending tickers from StockTwits for real-time retail sentiment and momentum.
Trending symbols · price + day gain
Congressional stock purchases from public STOCK Act disclosures. Cross-scan dedup prevents repeated ingestion of the same transaction.
Buys only · US tickers · 7-day pub window · txId dedup
Active prediction markets for stock catalysts — price targets, earnings beats, merger closes, FDA approvals, and S&P 500 inclusions. Two-phase scan: known symbols first, then any tickers discovered by other sources.
Public Gamma API · $5K total vol OR $1K 24h vol · event-level aggregation · two-phase scan
Raw mentions are grouped by ticker symbol. A symbol becomes a candidate when it appears ≥2 times from a single source, appears in ≥2 different sources, or comes from a high-value source (SEC Insider, Congress, Volume Spike, Options Flow) even as a single mention. Each source carries a weight that biases the aggregate score.
| Source | Weight |
|---|---|
| SEC Insider | 3.0 |
| Options Flow | 2.5 |
| Congress | 2.5 |
| Volume Spike | 2.5 |
| X / Twitter | 1.2 |
| SEC Filing | 1.0 |
| Polymarket | 2.0 |
| 1.0 | |
| StockTwits | 1.0 |
Each candidate is scored by AI using source weights, catalyst quality, novelty, and cross-source corroboration. Pure social signals (Reddit / StockTwits / Twitter only) are hard-capped at 50 — this is enforced programmatically regardless of what the AI returns. Only tickers with a verifiable catalyst source (SEC Insider or Congress) can score above 50. First-appearance tickers receive a +5–10 novelty boost; tickers seen 3+ times or older than 7 days receive a staleness penalty. Signal freshness is also tracked — stale consensus (median signal age ≥6 h) is excluded from the highest stage. This AI score reflects how strong the evidence is, not how much upside is left; Opportunity Score (see above) captures early-mover potential separately.
| Band | Meaning |
|---|---|
| 80–100 | Real catalyst + multi-source + insider/congress/options confirmation |
| 60–79 | Real catalyst + ≥2 sources, or strong insider/congress/options alone |
| 40–59 | Social buzz with catalyst indicators (unconfirmed) |
| 20–39 | Social-only signal, no verifiable catalyst |
| 0–19 | Likely noise or pump attempt |
Every candidate is checked against 13 statistical flags before scoring. Flags are split into effective flags (backed by ML as bearish predictors) and informational flags (detected but not counted toward the threshold). A ticker that triggers ≥3 effective flags is moved to Filtered status and quarantined. Exactly 2 flags triggers an additional AI edge-case assessment.
Score ≥40, multiple sources or novel ticker. Earliest detection point with highest alpha potential.
Score ≥45–50 with velocity or multi-source. Momentum is building but the move may have started.
Score ≥65–70 with broad, fresh social agreement or exchange-specific breakout patterns. Stale signals (median age ≥6 h) are excluded — the move may already be priced in.
Failed P&D check. Quarantined and visible in the Filtered tab.
Real catalyst + insider/options + multi-source corroboration (rare).
Real catalyst with ≥2 corroborating sources.
Interesting signal that needs further confirmation before acting.
No verifiable catalyst, pure hype, or P&D risk indicators.
SignalScope tracks the real-world performance of every signal it generates. Twice-daily price snapshots measure nominal returns at 1, 3, 7, and 30 days after detection. Tickers that undergo corporate actions (reverse splits, forward splits, mergers) during the tracking window are automatically detected via consecutive-snapshot analysis and excluded from performance statistics. This growing dataset trains a single LightGBM regression model (depth 2, 40 estimators) on 3-day forward returns across 308 engineered features — EWMA historical cross-products, P&D flag history, short-float and float-size interactions, and scan-level aggregates. Only about 13 features carry non-zero importance; the dominant predictor is the average signal strength across the scan, followed by the log of the interaction between a ticker's prior P&D reputation and its current P&D flag count, then scan size, log market cap, and the prior P&D × scan size interaction. The model is evaluated on 1-, 3-, and 7-day horizons; feature importance analysis identifies which factors drive accuracy and feeds back into AI score thresholds, stage assignments, and pump-and-dump detection — so the platform gets smarter with every scan.
SignalScope is for informational purposes only and does not constitute financial advice. Always do your own research before making any investment decisions.