Forecasting Engine Specific-Run Math Audit
This report documents the exact step-by-step mathematical calculations and feature vector scores executed during run 2026-10-31 02:17:07. All values below represent actual substituted numerical parameters generated for this specific forecast execution.
Section 1: Execution Audit & Trigger Headline Context
Daily Forecast Batch Execution (2026-10-31 02:17:07)
- NYMEX RBOB Futures & WTI Crude Spot Energy Commodity Benchmark Refresh (CME Group / NYMEX)
- NOAA National Weather Service Multi-Basin Severe Weather & Freeze Warning Ingestion (NOAA NWS Storm Alert)
- Executive Policy Feed & OPEC Weekend Open Price Gap Intelligence (Bloomberg Market Wire)
Section 2: Ingested Factor Score Vector (Exact Run Values)
Section 3: Step-by-Step Exponential Memory Decay Math for This Run
General Exponential Memory Decay Model Equation:
$$M_t = M_{t-1} \cdot e^{-\frac{\ln(2)}{t_{1/2}}} + S_t$$
Plugging in exact run decay parameters: decay constant \(\lambda = \frac{\ln(2)}{5.0} = 0.13863\text{ day}^{-1}\) and daily retention factor \(\gamma = e^{-0.13863} \approx 0.87055\).
Numerical Step-by-Step Shock Retention Schedule:
- • Day 0 (Initial Shock Target): $$M_0 = 0.1000$$
- • Day 1 Decayed Shock: $$M_1 = 0.1000 \times 0.87055 = 0.0871$$
- • Day 2 Decayed Shock: $$M_2 = 0.1000 \times (0.87055)^2 = 0.0758$$
- • Day 3 Decayed Shock: $$M_3 = 0.1000 \times (0.87055)^3 = 0.0660$$
- • Day 4 Decayed Shock: $$M_4 = 0.1000 \times (0.87055)^4 = 0.0574$$
- • Day 5 (Target Forecast Horizon): $$M_5 = 0.1000 \times 0.50000 = 0.0500$$
Section 4: Regional Metro Calibration Equations (Substituted Run Values)
$$\text{P}_{\text{National Wholesale}} = \$3.114 + (-\$0.009) = \$3.050\text{/gal}$$
$$\text{P}_{\text{Tulsa, OK Retail}} = \$3.803 + (+\$0.000) = \$3.803\text{/gal}$$
$$\text{P}_{\text{Newark, DE Retail}} = \$3.992 + (-\$0.013) = \$3.910\text{/gal}$$
$$\text{P}_{\text{Cincinnati, OH/KY}} = \$4.275 + (+\$0.000) = \$4.275\text{/gal}$$
$$\text{P}_{\text{Greenville, NC Retail}} = \$3.705 + (+\$0.023) = \$3.642\text{/gal}$$
$$\text{P}_{\text{Charlotte, NC Retail}} = \$3.447 + (-\$0.016) = \$3.362\text{/gal}$$
$$\text{P}_{\text{Port St. Lucie, FL Retail}} = \$3.969 + (-\$0.013) = \$3.880\text{/gal}$$
$$\text{P}_{\text{Chicago, IL/IN Retail}} = \$4.949 + (+\$0.000) = \$4.949\text{/gal}$$
$$\text{P}_{\text{Houston, TX Retail}} = \$3.671 + (+\$0.000) = \$3.671\text{/gal}$$
$$\text{P}_{\text{Los Angeles, CA Retail}} = \$6.021 + (+\$0.000) = \$6.021\text{/gal}$$
$$\text{P}_{\text{Denver, CO Retail}} = \$4.292 + (+\$0.000) = \$4.292\text{/gal}$$
$$\text{P}_{\text{New York, NY Retail}} = \$4.405 + (+\$0.000) = \$4.405\text{/gal}$$
$$\text{P}_{\text{Oakland, CA Retail}} = \$5.965 + (+\$0.000) = \$5.965\text{/gal}$$
Includes CA statutory CARB excise, Cap-and-Trade & LCFS fee overhead of $0.953/gal
$$\text{P}_{\text{SF Bay Area Region}} = \$6.003 + (-\$0.021) = \$5.875\text{/gal}$$
Includes CA statutory CARB excise, Cap-and-Trade & LCFS fee overhead of $0.953/gal
$$\text{DistillateCrack} = \$2.850 - \frac{\$75.00}{42.0} = \$0.742\text{/gal}, \quad \text{Crack}_{321} = \$0.685\text{/gal}$$
Experimental Work-In-Progress undergoing multi-week feedback loop empirical evaluation
Section 5: NOAA SPC-Style Quantitative & Narrative Synopsis
Issued: 2026-10-31 02:17:07Executive Forecast Summary
SUMMARY FOR RUN [2026-10-31 02:17:07]: Baseline daily batch market conditions prevail with minimal exogenous shocks. Ingested supply disruption S=0.10 and geopolitical risk G=0.15 yield a price pressure vector of ΔP=+0.02/gal. Primary trigger: 'Daily Forecast Batch Execution (2026-10-31 02:17:07)'. The standardized Ridge model calculates stable wholesale futures re-anchoring, with Day-5 residual event memory decaying from M₀=0.1000 down to M₅=0.0500.
Technical Discussion & Market Dynamics
Forecast Uncertainty & Counterfactual Catalysts
Section 6: Advanced Quantitative Feature & Physical Data Formulas
3-2-1 Refining Crack Spread Formula (Issue #169):
$$\text{Crack}_{321} (\$/\text{bbl}) = \frac{2 \times (P_{\text{RBOB}} \times 42) + 1 \times (P_{\text{HO}} \times 42) - 3 \times P_{\text{WTI}}}{3}$$
Models refiner yield switching between gasoline and distillate heating oil.
Stacking Ensemble Quantile Bands (Issue #170):
$$P_{10} = P_{50} - 1.2815\sigma, \quad P_{90} = P_{50} + 1.2815\sigma$$
Computes 80% probabilistic confidence intervals using RidgeCV meta-learner variance.
Dynamic Volatility-Gated Persistence Blending (Issue #214):
$$\lambda_{\text{vol}} = \frac{1}{1 + e^{-200.0 \cdot (\sigma_{14\text{d}} - 0.015)}}, \quad \hat{y}_{t+5} = \lambda_{\text{vol}} \hat{y}_{\text{model}} + (1 - \lambda_{\text{vol}}) y_t$$
Blends model forecasts into naive persistence during low-volatility plateaus while preserving 100% shock reactivity.
Empirical Residual 95% Confidence Intervals (Issue #214):
$$\text{CI}_{95\%} = \hat{y}_{t+5} \pm 1.96 \cdot \sigma_{\text{residual, 30d}}(r)$$
Derives dynamic 95% confidence bands from rolling 30-day regional residual standard errors.
USGS 3D Hypocentral Attenuation & Ground Shaking (Issue #55):
$$R = \sqrt{d^2 + h^2}, \quad w(R) = \frac{1}{1 + (R/35)^2}, \quad I = 10^{M - M_{\text{base}}} \times w(R)$$
Quantifies facility-level peak ground shaking and pipeline shutoff risk indices across 5 corridors.
USGS Hydrological Barge Bottleneck Index (Issue #56):
$$\text{Index}_{\text{barge}} = \max\left(0, \min\left(1, \frac{\text{Gage}_{\text{threshold}} - \text{Gage}_t}{\text{Gage}_{\text{threshold}} - \text{Gage}_{\text{min}}}\right)\right)$$
Tracks low-water navigation draft restrictions at Memphis & Cairo confluence throttling barge throughput.
Multi-Feed AQI Standardized Flaring Outage Z-Score (Issue #54):
$$Z_{\text{PM2.5}} = \frac{\text{PM2.5}_t - \mu_{30\text{d}}}{\sigma_{30\text{d}}}, \quad Z_{\text{SO2}} = \frac{\text{SO2}_t - \mu_{30\text{d}}}{\sigma_{30\text{d}}}$$
Detects emergency refinery flaring when $Z_{\text{PM2.5}} \ge 3.5 \land Z_{\text{SO2}} \ge 2.5$ with 12-24h lead time over news.
EPA Ozone & Statutory Seasonal RVP Compliance Surcharge (Issue #73):
$$\text{Surcharge}_{\text{RVP}} = \Delta \text{Spread}_{\text{Summer Blend}} + 0.040 \cdot \mathbf{1}_{\text{AQI}_{\text{O3}} \ge 101}$$
Applies statutory summer-blend RVP constraints (CARB 7.0 psi, EPA 7.8 psi, Conventional 9.0 psi) and Ozone Action Day fees.
U.S. Census Commuter Inelastic Demand Score (Issue #75):
$$\text{Score}_{\text{inelastic}} = \frac{\text{DriveAlone} + \text{Carpool}}{\text{TotalCommuters}} \times (1 - \text{TransitIndex})$$
Calibrates retail pump price pass-through speed and baseline rack spreads from ACS commuting tables.
Treasury 10Y-2Y Term Spread & Momentum Delta (Issue #66):
$$\text{Spread}_{10\text{Y}-2\text{Y}} = Y_{10\text{Y}} - Y_{2\text{Y}}, \quad \Delta \text{Spread}_{5\text{d}} = \text{Spread}_t - \text{Spread}_{t-5}$$
Captures leading macroeconomic expansion/recession demand signals and real TIPS inventory carry costs.
Qlib Symbolic Alpha Information Coefficient (Issue #127):
$$IC_t = \text{Corr}(f_t, r_{t+h}), \quad IC_{IR} = \frac{\mu(IC)}{\sigma(IC)}$$
Evaluates AST-parsed symbolic alpha factor formulas with non-lookahead point-in-time calculation rules.
Multi-Horizon Scoreboard Performance Accuracy (Issue #209):
$$\text{MAE}_H = \frac{1}{N_H} \sum_{i=1}^{N_H} |\hat{y}_{i, H} - y_{i, H}|, \quad H \in [1\text{d}, 2\text{d}, 3\text{d}, 4\text{d}, 5\text{d}]$$
Tracks continuous rolling out-of-time accuracy breakdowns across all discrete forecast horizons.