How every signal
gets made
From data ingestion at 4:00 AM ET to delivery at 7:00 AM ET — a full breakdown of our pipeline, scoring model, confidence score methodology, and the explicit limits of what our signals claim.
Overview
The pipeline at a glance
4:00 AM ET
Data ingestion begins
Pipeline activates, pulling structured feeds across all six source categories.
4:00–6:30 AM
Event detection & deduplication
Candidate events are extracted, cross-referenced across sources, and deduplicated so the same event is not counted twice.
6:00–6:45 AM
AI scoring
Each deduplicated event is scored across four dimensions. Directional bias and confidence score are produced.
6:45–7:00 AM
Ranking, assembly & delivery
Events are ranked by estimated price impact. The brief is assembled, human-readable context is generated, and the edition is delivered.
7:00 AM ET
Subscriber delivery
Pro and Trader subscribers receive the edition. Free plan subscribers receive it at 9:00 AM ET with a 2-hour delay.
Data ingestion
Each weekday starting at 4:00 AM ET, our pipeline ingests structured feeds across six source categories. Sources are deduplicated and cross-referenced so the same event is not counted twice — a critical step, since the same earnings report may appear across SEC EDGAR, wire services, and a dozen financial news outlets within minutes.
SEC & Regulatory Filings
Earnings releases (10-Q, 10-K), 8-K event filings, Form 4 insider transaction disclosures, and proxy filings. These are the authoritative primary sources for corporate events — any delay in ingesting them is a delay in signal generation.
Central Bank & Government Data
Federal Reserve meeting statements, FOMC minutes, Fed official speeches, and US government data releases including CPI (Consumer Price Index), NFP (Non-Farm Payrolls), PMI (Purchasing Managers Index), GDP revisions, and Treasury yield data.
Wire & Financial News Services
Structured news feeds from financial wire services, cross-referenced against multiple outlets to distinguish primary events from commentary. Opinion pieces and analyst notes are flagged separately from factual event reporting and weighted accordingly.
On-Chain Blockchain Indexers
Real-time blockchain data for crypto signals: Bitcoin and Ethereum ETF inflow/outflow data, exchange reserve movements (coins moving to/from exchange wallets), large wallet transaction monitoring, and stablecoin supply shifts that historically precede directional price moves.
Pre-Market Price & Options Flow
Pre-market equity prices (4:00–9:30 AM ET), overnight futures data, and unusual options activity. Options flow data — particularly large block trades, high put/call ratio shifts, and near-expiry volume spikes — provides forward-looking market sentiment that supplements news-based signals.
Social Sentiment Aggregators
Aggregated directional sentiment across financial social platforms, weighted heavily by account credibility and historical signal reliability rather than raw volume. Raw social sentiment is treated as a weak corroborating signal only — it never drives a directional call independently.
Market data is sourced via Finlight and supplemented with direct exchange and blockchain data feeds.
AI scoring
Each deduplicated candidate event is scored across four dimensions. The combination produces a directional call (bullish, bearish, or neutral) and a confidence score from 0–100%. No single dimension determines the output — all four must align for a high confidence score to be produced.
Source agreement
30%How many independent sources confirm the same event and the same directional interpretation. A single source reporting an event scores low on this dimension regardless of the source's credibility. Convergence across structurally independent sources — for example, an 8-K filing, a wire report, and pre-market price action all pointing the same direction — scores high.
Historical pattern match
25%How comparable events have moved the affected asset over 1-day and 5-day windows historically. The model maintains a structured database of past event-outcome pairs by event type (earnings miss, Fed rate hold, ETF outflow above threshold, etc.) and asset. A strong pattern match means the event type has a well-established historical directional distribution for this asset class.
Sentiment signal strength
25%The magnitude and consistency of directional language around the event, not just its polarity. A weak bearish signal with mild language across three sources scores differently from a strong bearish signal with high-conviction language and consistent framing. This dimension penalizes ambiguity.
Pre-market price confirmation
20%Whether early price action — equity pre-market trading, futures movement, or crypto spot price — already validates the thesis before the brief is assembled. A bullish signal where the asset is already up 4% pre-market scores higher on confirmation than one where the price has not yet moved. This factor is updated as close to the 7 AM delivery window as possible.
How to read confidence scores
The confidence score reflects how strongly the model's evidence base aligns with its directional call. It is not a probability of profit, a guarantee of accuracy, or a prediction about long-term performance. It is a measure of internal consistency across the four scoring dimensions for this specific event.
Strong multi-factor alignment. Multiple independent sources confirm the event and direction, historical pattern is clear, sentiment is consistent, and pre-market price action validates the thesis. These signals have historically shown the strongest directional alignment.
Most factors align but at least one is mixed or inconclusive. The event is notable and directional evidence is present, but not without ambiguity. Use alongside your own analysis.
Notable event with mixed signals. Published for awareness, not conviction. Treat these as contextual information rather than actionable trade triggers. Do not use sub-65% signals as the primary basis for a position.
Events that do not reach 55% confidence are not included in the brief. Our ranking model filters aggressively — a signal that does not meet this threshold adds noise, not value.
Ranking, assembly & delivery
Ranking. Scored events are ranked by estimated price impact — a composite of confidence score, asset liquidity, sector breadth of the event, and historical magnitude of comparable events. Macro events that affect multiple asset classes are ranked higher than single-stock events at equivalent confidence scores.
Plan-tier filtering. The top signals are delivered based on your plan: Free (3 signals, 2-hour delay), Pro (10+ signals, 7:00 AM ET), Trader (10+ signals plus Telegram alerts for intraday events). Higher-tier plans receive more signals and the real-time advantage of early delivery — the underlying signals are the same, not a watered-down version for lower tiers.
Assembly. For each ranked signal, human-readable context is generated for the “Why it matters” field — a concise explanation of the mechanism by which the event is expected to affect price, with key levels to watch and any relevant cross-asset context. This field is generated to inform, not to recommend.
Delivery. The complete edition is delivered at 7:00 AM ET — approximately two and a half hours before the US market opens at 9:30 AM. On US market holidays, no edition is sent. Editions are self-contained and clearly dated so missed days can be caught up without confusion.
What Alphora Signals does not claim
Signals are informational context, not instructions. Past signal performance does not guarantee future results. Always conduct your own research and consult a qualified financial professional before making any investment decision. See our Terms of Service for the full scope of service.
See it in action
The sample edition shows exactly what arrives in your inbox every morning — real format, illustrative data.