How Paradexa forms a bias

Each economic release maps to a macro factor (inflation, growth, labor, monetary policy, rates, currency, risk sentiment, energy, liquidity). For released events we compute a normalized surprise (actual − forecast) / threshold, clipped to [-2, +2], then weighted by event importance.

Upcoming events do not contribute a surprise — they only raise event risk and reduce confidence in dependent assets.

Each asset carries a sensitivity vector to macro factors — for example, gold is negatively sensitive to rate expectations and USD strength. Contributions equal factor score × sensitivity × factor weight. The total is mapped to a bias label.

Score ≥ +1.2 → Strong Bullish · ≥ +0.4 → Bullish · between → Neutral · ≤ −0.4 → Bearish · ≤ −1.2 → Strong Bearish. When bullish and bearish contributions are of comparable size, the label becomes Mixed rather than being forced to a side.

Confidence starts at 70 and moves with data quality: released, corroborating data increases it; missing/delayed data, conflict between factors, upcoming high-impact catalysts, staleness, or too few observations reduce it. Confidence is capped at 95 and never claims certainty.

Before a release, we show scenario analysis (hotter / in line / cooler) so you can prepare, and we clearly mark the bias as conditional. After a release, we replace scenarios with actual-vs-forecast, surprise magnitude, and which assets received positive or negative pressure.

Historical macro–asset relationships are regime dependent. Good growth data can support equities in one regime and hurt them in another (if it raises rate expectations too much). The sensitivity matrix is editable and versioned in the database so historical biases remain reproducible.

We never silently fill missing values with zero. If a provider is offline, we show a stale-data warning and reduce confidence on affected assets. We never label data "real-time" unless the underlying provider actually delivers real-time updates.

Paradexa provides educational market information and probabilistic analysis, not personalized investment advice. Verify data with original sources before making financial decisions.