How We Score Anomalies

Signals we observe

• Open Interest (OI) change vs rolling window
• Volume spike vs typical time-of-day baseline
• Implied Volatility (IV) shock/inflection
• VWAP divergence vs recent price path

Normalization & z-scores

We compute feature deltas relative to rolling means/medians with dispersion estimates, then convert to comparable scales (e.g., z-like scores). This avoids over-emphasizing noisy symbols.

Composite anomaly score

A weighted combination of the standardized features forms the headline score (0–100). Higher means “more unusual relative to recent context,” not “buy/sell.”

• Typical weights (illustrative): OI 30%, Volume 30%, IV 25%, VWAP 15%
• Caps and decay prevent one-off prints from dominating
• Time-of-day adjustment reduces opening/closing noise

Important caveats

• Scores reflect relative unusualness, not profitability
• Market regimes shift; baselines adapt but are not perfect
• Use judgment; TradingRadar is educational only