The Single-Source Trap
You see a congressional trade alert: Nancy Pelosi just bought NVDA calls. You buy.
Three days later, you're down 8%. What happened?
Here's the problem: single-source trading is gambling dressed up as strategy.
That congressional trade might have been a hedge against an existing position. It might have been filed 45 days after execution. Or maybe the broader market was about to tank, and no amount of insider knowledge could overcome macro headwinds.
Single signals fail because markets are complex systems. No single data source—no matter how "smart" the money behind it—captures that complexity.
What Is Signal Fusion?
Signal fusion is the practice of combining multiple independent data sources into a unified confidence score. Instead of asking "what is Congress doing?" you ask "what are Congress, hedge funds, dark pools, and technical indicators all saying about this stock?"
When multiple independent sources point the same direction, you have something powerful: confluence.
Think of it like courtroom evidence. One eyewitness? Unreliable. DNA evidence alone? Could be contaminated. But DNA + fingerprints + eyewitness + motive + opportunity? That's a conviction.
Signal fusion works the same way.
How AlphaSignal Fuses 14 Data Sources
Our system ingests signals from 14 distinct sources:
| Category | What It Tracks |
|---|---|
| Congressional Trades | STOCK Act disclosures from elected officials |
| 13F Institutional | Quarterly hedge fund position changes |
| Dark Pool Activity | Block trades on alternative trading systems |
| Insider Transactions | CEO/CFO/Director buys and sells |
| Short Interest | FINRA bi-monthly settlement data |
| Short Volume | Daily short sale volume |
| Options Flow | Unusual sweeps and block activity |
| FTD Data | SEC fails-to-deliver spikes |
| Form 8-K Filings | Material corporate events |
| WARN Act | Layoff notices (60-day advance warning) |
| Economic Events | CPI, FOMC, NFP impact modeling |
| Technical Analysis | RSI, MACD, volume extremes, patterns |
| Fundamental Metrics | P/E, FCF yield, ROE |
| ML Forecasts | LSTM and ensemble predictions |
But raw signals aren't useful. The magic is in how we combine them.
The Math Behind Confidence Scores
Every signal that enters our system carries two pieces of information:
- Direction — Is it bullish or bearish?
- Historical Win Rate — How often has this exact signal type been right over the past 30 days?
We calculate a base confidence score using weighted averages:
Base Confidence = Σ(Win Rate × Adaptive Weight) / Σ(Adaptive Weight)
Signals that have been performing well recently get higher weights. Signals that have been underperforming get downweighted. This happens automatically, every day.
Why Diversity Matters: The Confluence Boost
Here's where it gets interesting. We don't just average signals—we reward diversity.
If you have three bullish signals, but they're all technical indicators, that's weaker than having one congressional trade, one dark pool signal, and one 13F filing. Why? Because technical indicators can all be wrong for the same reason (fake breakout, algo manipulation). But Congress, hedge funds, and dark pools failing simultaneously? That's rare.
Our confluence boost rewards this:
| Signals | Base Boost | Type Diversity | Source Diversity | Max Total |
|---|---|---|---|---|
| 1 signal | +0 | — | — | 0 |
| 2 signals | +3 | +1 | +1 | +5 |
| 3 signals | +7 | +3 | +3 | +13 |
| 4+ signals | +10 | +5 | +3 | +18 |
A stock with four diverse signals agreeing can get up to an 18-point confidence boost. That's the difference between "maybe" and "high conviction."
Real Example: How Fusion Creates Edge
Let's say AAPL has the following active signals:
| Source | Signal | 30-Day Win Rate |
|---|---|---|
| Congress | Bought | 68% |
| Dark Pool | Accumulation | 72% |
| 13F | New Position | 65% |
Base Confidence: 68.5%
Confluence Boost:
- 3 signals = +7 points
- 3 different types = +3 points
- 3 different sources = +3 points
- Total boost: +13 points
Final Score: 82 (Strong)
Now compare to a stock with only one signal—even if that signal has a 72% win rate, it caps out around 72. No boost. No confluence. No edge multiplication.
The Threshold That Keeps You Out of Bad Trades
We don't show you every signal. Most never reach your screen.
To become an actionable alert, a symbol must meet two conditions:
- Minimum Confidence: 55+ — Below this, the data is too ambiguous
- Minimum Edge: 8+ points — The bullish score must exceed the bearish score by at least 8 points
This means if AAPL has an 82 bullish score but also a 76 bearish score (only 6-point edge), we won't alert you. The signal is conflicted.
We'd rather miss a trade than put you in a coin flip.
Adaptive Learning: The Weights Evolve
Markets change. What worked in 2023 might not work in 2024.
Our system recalculates weights daily based on 30-day rolling performance:
| Win Rate | Weight Assigned |
|---|---|
| ≥70% | 0.90 (heavy) |
| ≥65% | 0.80 |
| ≥60% | 0.70 |
| ≥55% | 0.60 |
| ≥50% | 0.50 |
| <50% | 0.40 (light) |
If congressional trades start underperforming, they automatically get downweighted. If dark pool signals go on a hot streak, they get amplified. No manual intervention needed.
This is what "adaptive" means—the system learns from its own outcomes.
Why Single-Source Services Can't Compete
Most trading alert services give you one thing:
- Congressional trade trackers show you congressional trades
- Options flow scanners show you options flow
- 13F aggregators show you 13F filings
But they can't answer the question that matters: "Should I actually trade this?"
You're left to synthesize the data yourself. And that synthesis—the weighting, the conflict resolution, the regime awareness, the performance tracking—is the hard part.
Signal fusion does that synthesis for you, algorithmically, with full transparency into how scores are calculated.
The Bottom Line
Single-source trading is like diagnosing a patient with only a blood pressure reading. You might get lucky. But you're flying blind.
Signal fusion combines the evidence:
- What is Congress doing?
- What are institutions doing?
- What is dark pool volume showing?
- What are the technicals saying?
- Do they agree?
When they agree—and only when they agree—you get an alert.
That's not a tip. That's a thesis with receipts.