Technology

Signal Intelligence Engine

The layer between raw market data and trading decisions. Our Signal Intelligence Engine analyzes, filters, scores, and structures market opportunities before they ever reach you — or Nova.

The Problem

Most trading systems react to noise.

Conventional alert platforms fire notifications at every RSI cross, every volume spike, every breakout attempt. The result is a high-volume stream of low-conviction triggers that overwhelm rather than inform.

Trading bots that consume these raw signals inherit the same problem — acting on noisy inputs with no intermediate layer of judgment. The result is overtrading, inconsistent execution, and difficulty distinguishing signal from noise.

We built the Signal Intelligence Engine to solve this. It sits between raw market data and action — refining opportunities before they reach users or automation.

How It Works

Six stages. One clean signal.

01

Data Ingestion

The engine continuously scans 100+ symbols across crypto and US stocks, fetching 120 bars of daily price history per symbol every 2 minutes. Multiple data sources ensure broad coverage and redundancy.

Sources: Alpaca Market Data API for stocks and crypto OHLCV bars.
02

Strategy Analysis

Each symbol is evaluated by 8 independent strategy filters running in parallel. Each strategy looks for specific market conditions — trend alignment, reversal patterns, volatility compression, momentum breakouts, and more.

Strategies: Trend, Mean Reversion, Double Reversal, Gap Continuation, Volatility Squeeze, Ichimoku, Supertrend, Momentum Breakout.
03

Confidence Scoring

Strategies that detect a potential opportunity produce a confidence score between 0 and 1. Only signals exceeding the confidence threshold are promoted. When multiple strategies agree, confluence increases the signal's weight.

Minimum threshold: 0.60 confidence. Multi-strategy confluence adds additional conviction.
04

Quality Validation

Each signal is assessed for data quality — bar count, data latency, volume consistency. Signals generated from incomplete or stale data are penalized or rejected before reaching users.

Quality score: 0–1 scale. Penalizes low bar counts, high latency, zero-volume bars.
05

Signal Structuring

Validated signals are structured into a clean, standardized format: entry price, stop-loss, take-profit targets, confidence score, quality rating, strategy attribution, and a plain-language rationale.

Schema: Clean Signal Contract v1.0.0 — validated by both producer and consumer.
06

Delivery

Clean signals are published to the Horizon dashboard for subscribers and delivered to Nova via HMAC-signed HTTP for automated execution. Every delivery is logged for audit.

Transport: HMAC-SHA256 signed, idempotent, with replay protection and retry logic.

Every signal is scored and structured.

Confidence Score0 – 1

How strongly the strategies agree on this opportunity. Higher scores indicate stronger multi-strategy confluence.

Quality Score0 – 1

Data quality assessment. Accounts for bar completeness, data latency, and volume consistency.

Entry + LevelsStructured

Entry price, stop-loss, and take-profit targets are computed and included with every signal.

RationalePlain text

Every signal includes a human-readable explanation of why it was generated and which factors contributed.

Built for humans and machines.

Clean signals are designed to be actionable by both manual traders and our Nova automation engine. Same intelligence, your choice of control.