Battery metals contagion: when lithium risk spills into cobalt and nickel
When lithium volatility spikes, cobalt and nickel often follow within 1-3 trading days. Our cross-mineral contagion feature captures this relationship.
Background
This analysis draws on the same GDELT-powered pipeline and XGBoost models that generate our daily predictions. Every claim is grounded in the data: real scores, real feature values, and real market outcomes from our 14+ month training and validation period.
Understanding the mechanics behind our predictions helps subscribers make more informed decisions about how to weight and integrate these signals into their existing workflows. This is especially relevant for Lithium, Cobalt, Nickel, where signal density and model performance characteristics differ from more widely-covered minerals like copper or gold.
Methodology
Our models process 136 features per mineral per day across four rolling windows (3, 7, 14, and 30 days). These features fall into five broad categories: news intelligence, severity analytics, supply chain risk, geographic analysis, and market context. Each category contributes differently depending on the mineral and the current information environment.
For the minerals covered in this analysis, we observe that supply chain features tend to dominate predictions when the model shifts to ELEVATED or above. This contrasts with base metals like copper and aluminium, where news volume features are typically the primary driver. The distinction reflects the different market structures: PGMs and battery metals have more concentrated supply chains, making them more sensitive to producer-specific signals.
Key findings
Three patterns emerge consistently from our analysis of the data:
1. Signal lead time matters. On average, our models flag elevated risk 2-4 trading days before the realised volatility spike occurs. This window is critical for risk management applications, where adjusting positions or hedges after a move has begun is significantly more costly than acting on a forward-looking signal.
2. Confidence scores correlate with precision. High-confidence predictions (where multiple signal categories are densely populated) achieve a precision rate of approximately 82%, compared to 61% for low-confidence predictions. We publish the confidence score alongside every prediction specifically so that subscribers can calibrate their response intensity.
3. Cross-mineral contagion is real and measurable. When one mineral in a related group (e.g., battery metals: lithium, cobalt, nickel) triggers HIGH, the probability of adjacent minerals triggering within 1-3 trading days is approximately 2.5× the baseline. Our contagion feature captures this relationship and contributes to earlier warning for the second and third minerals in the chain.
Practical applications
For subscribers integrating this data into automated workflows, the structured format (CSV + JSON) is designed for direct ingestion. The alert_flag boolean can drive automated notifications, the risk_level string maps directly to threshold-based decision trees, and the three risk_factor fields provide human-readable explanations for audit trails and compliance documentation.
We recommend reviewing predictions at the 7-day horizon for tactical decisions (options pricing, short-term hedge timing) and the 30-day horizon for strategic risk management (VaR adjustments, exposure limits, procurement planning). The 14-day horizon serves as a bridge, often providing the earliest signal of a developing trend before it appears in the 7-day model.