Signal Lifecycle & Scoring Pipeline

Every trade in TRADEOS.tech begins as a raw signal event and passes through a multi-stage pipeline before it can result in an order. Understanding this pipeline explains why TRADEOS.tech produces fewer, higher-quality trades rather than acting on every indicator firing.

Overview

A signal travels through six sequential stages before becoming an execution candidate:

Raw Event → Enrichment → Quality Scoring → Confluence Gate → Composite Score Gate → Feasibility → Execution

Each stage can reject or modify the signal. A signal that makes it through all stages to execution has survived multiple independent quality checks.

Stage 1: Raw Signal Generation

Raw signals are generated by individual alpha models monitoring market data streams. Each raw signal carries:

• Symbol — the instrument being signaled
• Direction — BUY or SELL
• Base confidence — the model's initial certainty estimate (0.0–1.0)
• Signal type — which alpha category generated it (momentum, order_flow, on_chain, etc.)
• Timestamp — when the condition was detected

At this stage, no quality filtering has occurred. The signal is just an observation.

Stage 2: Enrichment

Raw signals are enriched with contextual data before scoring:

• OHLCV context: volatility regime (low / normal / high / extreme), volume confirmation, trend alignment — each adjusts base confidence and sizing upward or downward
• Regime context: the current HMM regime state and BOCD change-point probability are attached
• VPIN health: adverse selection risk measure from order book analysis
• Hawkes OFI: current order flow imbalance from Hawkes process model

Enrichment transforms a raw signal into a context-aware signal with adjusted confidence and a sizing pre-multiplier.

Stage 3: Quality Scoring

The quality scorer computes a composite quality score (0.0–1.0) from multiple dimensions:

• IC multiplier — how well has this signal type historically predicted outcomes?
• Calibration score — does the stated confidence reflect actual win rates?
• Regime fit — is the current regime one where this signal type performs?
• Confluence velocity — is this signal type arriving in a meaningful burst or thin trickle?
• Sample sufficiency — is there enough live-trading history to trust quality estimates?

Signals below the minimum quality threshold for the active strategy profile are rejected here before reaching the confluence gate.

Stage 4: Confluence Gate

The confluence gate requires that multiple independent sub-models agree before a signal advances. Each signal event is scored against all active sub-models, and the gate checks:

1. How many independent sub-models are producing the same direction?
2. Is there sufficient directional agreement (score above threshold)?
3. Are any opposing signals present that need resolution?

This stage is where most noise is filtered. A single indicator firing in isolation never generates a trade — the signal must be corroborated.

Stage 5: Composite Score Gate

Signals that pass confluence are evaluated against a tiered composite score gate. The composite score combines quality, confidence, regime fit, and IC into a single 0.0–1.0 number. This score maps to a sizing tier:

Score Range	Tier	Effect
Below minimum	Blocked	Signal rejected — insufficient quality
Low	Half	Reduced position size
Mid	Full	Standard position size
High	Bonus	Elevated position size (for exceptional setups)

Only signals in a qualifying tier proceed to execution. The minimum score threshold and tier boundaries vary by strategy profile.

Stage 6: Feasibility Check

The final pre-execution gate validates that the trade is feasible given current portfolio state:

• Is there available capital for this position?
• Would this trade exceed concentration limits?
• Is the symbol already at maximum exposure?
• Does the risk/reward calculation meet the minimum R threshold?

Feasibility rejection is not a signal quality issue — it is a portfolio management constraint.

Signal Lineage

Every signal that completes the pipeline carries a lineage record — a timestamped log of every stage it passed through, the scores at each stage, and any modifications applied. This lineage is stored with the trade record and is available for inspection in the dashboard and audit trail.

Signal lineage is the mechanism that makes TRADEOS.tech's decision-making fully auditable: for any trade, you can trace exactly why the system decided to enter, what adjustments were applied, and what the scores were at each stage.

Decay and Expiry

Signals have a time-to-live. A signal generated but not executed within its validity window is discarded rather than executed stale. Different signal categories have different decay rates:

• Order flow signals decay within minutes — microstructure windows are brief
• Momentum signals have longer validity — trends persist longer than order flow bursts
• On-chain signals have the longest validity — blockchain-level conditions change slowly

This prevents stale signals from being triggered at the wrong time due to pipeline latency.