The 5 layers explained
Each TradeVelocity signal is scored by an AI engine across five independent layers. The model only emits when enough of them agree, and the grade of the signal reflects how strongly they agreed.
The five layers are deliberately independent — they look at different things and can disagree. When they line up, that's confluence. When they don't, the engine stays quiet.
Layer 1: Market Structure
Reads higher-timeframe price action — are we making higher highs and higher lows (uptrend), lower highs and lower lows (downtrend), or chopping sideways (range)?
Range markets get less weight in the score. Chop is the enemy of confluence — most failed signals trace back to range-bound conditions where direction is meaningless.
Layer 2: Trend
Combines moving-average alignment on a working timeframe with a separate trend-strength check. The model wants to see both: the structure of a trend AND evidence that the trend is being driven, not drifting.
A market that "looks bullish" without volume + momentum behind it is a setup for a fade. The AI rejects those.
Layer 3: Momentum
Looks for momentum that's building, not exhausted. Extreme readings mean the move has already happened — entering then is buying tops or selling bottoms. The model favors the middle of the momentum range, where there's room for the trade to work.
Layer 4: Volume
Two checks: short-term volume relative to recent baseline (is participation expanding?), and the directional pressure inside that volume. Real moves come with confirming volume. Chop moves don't. The AI scores both.
Layer 5: Volatility
Combines a range-expansion measure with compression detection. The model wants the move to be releasing energy, not still coiled or already exhausted.
How layers combine into a grade
The AI weighs each layer plus its sub-indicators and produces a confidence score. Higher grades require more layers in stronger agreement; lower grades fire when the model has partial conviction. The four grades:
| Grade | Meaning |
|---|---|
| A+ | Highest confidence — every layer aligned |
| A | Strong — most layers + indicators agree |
| B+ | Speculative-moderate — partial confluence |
| B | Speculative — minimum threshold met |
The exact weighting is part of the model and tuned continuously by backtest. Subscribers see the grade — the math behind the grade is the engine's job, not the trader's.
Why we cap confidence at 90%
No model is 100% sure. A score that high would imply mathematical certainty about a probabilistic system. The AI deliberately leaves a humility ceiling — it acknowledges its own limits.
That cap is also room for future model improvements. As we collect more data and refine the layer weights, signals can improve without the score becoming meaningless.
