# Executive Summary

## Headline

The strongest stable lead-time range for G4 GM2 is 5 to 6 months in rolling validation, with the locked final reporting model set to GM2-only lag 5. Simple lag correlation peaks at 4 months for both WTI and Brent.

## GM2-Only Performance

- WTI YoY, GM2-only lag 5: rolling RMSE 31.628, MAE 22.295, directional accuracy 49.8%.
- Brent YoY, GM2-only lag 5: rolling RMSE 32.358, MAE 23.990, directional accuracy 49.8%.

## Comparative Inventory Role

Comparative inventory does not improve primary Oil_YoY rolling RMSE or MAE versus GM2-only by the 5% rule. Its useful role is diagnostic: it helps explain whether oil is rich or cheap versus the liquidity-implied path, especially when combined with regime labels.

## Where The Model Fails

The model is weakest around abrupt geopolitical, pandemic, financial-crisis, shale-cycle, and policy shocks. It also does not directly model OPEC behavior, spare capacity, refining margins, curve structure, global inventories, sanctions, or shipping disruptions.

## Current Signal

As of 2026-05, WTI YoY is 64.288. The locked GM2-only model implies 38.180, leaving a residual of 26.108. Inventory state: inventories are near normal versus history, so CI is not sending a strong surplus/deficit signal.

## What To Watch Next

- Whether G4 GM2 YoY keeps accelerating or rolls over, because the locked model reads that impulse with a 5-month lead.
- Whether comparative inventory stays in surplus or deficit, because that frames residual risk around the liquidity path.
- Whether the current period resembles a shock regime where the simple liquidity relationship is more likely to break.
- Whether Brent and WTI residuals diverge, which can signal location-specific physical-market stress rather than broad liquidity momentum.