Marlin Intelligence methodology
How the Marlin Score works: 15 WTI factors, scoring rules, weights, regime thresholds, confidence, missing data and the limits of the AI layer.
The Marlin Score is a deterministic summary of 15 WTI factors. Each factor gets a score from -10 to +10 by a fixed rule. The score is their weighted sum, rescaled to -100 to +100, and it maps to one of five regime bands. It describes current conditions. It is not a forecast, a price target or a trade signal.
How is the Marlin Score calculated?
- Collect. Each factor reads one data series (two for the Brent-WTI spread) from a named source, with the period the value describes.
- Score each factor. The factor's change or level is multiplied by a fixed scale and clamped to -10 to +10. A positive score means the reading is supportive for WTI; a negative score means it is pressuring.
- Combine. Marlin Score = Σ(weight × factor score) ÷ Σ(weight × 10) × 100, over factors that have data, rounded to a whole number.
- Label. The rounded score falls into one of the regime bands below.
Only the numbers are deterministic. The same inputs always give the same score, and no AI is involved in any step above.
What do the regime bands mean?
The dashboard calls this the Current Market Regime: a label for the balance of factors today.
| Marlin Score | Regime |
|---|---|
| +60 to +100 | Bullish |
| +20 to +59 | Moderately Bullish |
| -19 to +19 | Neutral |
| -59 to -20 | Moderately Bearish |
| -100 to -60 | Bearish |
Which factors are scored, and how?
Weights sum to 14.8. A factor whose absolute score is below 1 reads as neutral.
| Factor | Weight | Rule | Source |
|---|---|---|---|
| WTI price structure | 1.0 | Distance of WTI spot from its 50-session average: +1.5 points per 1% above. | EIA, Cushing OK WTI spot price FOB (RWTC) |
| Brent | 1.0 | Daily % change in Brent spot: +3 points per 1% rise. | EIA, Europe Brent spot price FOB (RBRTE) |
| Brent-WTI spread | 0.5 | Change in Brent minus WTI: +5 points per $1 of widening (a wider spread makes US barrels more competitive for export). | EIA, Europe Brent spot price FOB (RBRTE) |
| US dollar (broad index) | 1.0 | Daily % change in the Fed's broad dollar index: -6 points per 1% rise. | Federal Reserve H.10 via FRED (DTWEXBGS) |
| US Treasury yields | 0.6 | Daily change in the 10-year yield: -0.4 points per basis point rise (tighter financial conditions). | Federal Reserve H.15 via FRED (DGS10) |
| US crude inventories | 1.5 | Weekly change in US commercial crude stocks: +1.2 points per 1 million barrel draw. | EIA Weekly Petroleum Status Report (WCESTUS1) |
| Cushing inventories | 1.0 | Weekly change in Cushing stocks (the WTI delivery hub): +3 points per 1 million barrel draw. | EIA Weekly Petroleum Status Report (W_EPC0_SAX_YCUOK_MBBL) |
| US crude production | 1.0 | Weekly change in US output: -4 points per 100 kb/d rise. | EIA Weekly Petroleum Status Report (WCRFPUS2) |
| US refinery utilization | 0.8 | Weekly change in utilization: +2.5 points per percentage-point rise (more crude demand from refiners). | EIA Weekly Petroleum Status Report (WPULEUS3) |
| OPEC+ supply | 1.5 | Monthly change in OPEC+ crude output: -2.5 points per 100 kb/d rise. | OPEC Monthly Oil Market Report (secondary sources), entered by the Marlin desk |
| China demand indicators | 1.0 | China crude imports y/y: +0.8 points per 1% above last year. | General Administration of Customs of China, entered by the Marlin desk |
| Geopolitical risk | 1.2 | GPR index level vs its 100 baseline: +1 point per 6 index points above. | Caldara & Iacoviello Geopolitical Risk Index (monthly) |
| Strategic Petroleum Reserve | 0.4 | Weekly change in SPR stocks: +2 points per 1 million barrel refill (government buying). | EIA Weekly Petroleum Status Report (WCSSTUS1) |
| Futures curve / market structure | 1.5 | Front-month minus second-month WTI: +6 points per $1 of backwardation, negative in contango. | EIA, NYMEX WTI futures contracts 1 and 2 (RCLC1, RCLC2) |
| Risk sentiment / equities | 0.8 | Daily % change in the S&P 500: +3 points per 1% rise. | S&P 500 index close (licensed market data provider) |
What happens when data is missing or old?
A factor without a current value shows Data unavailable and is left out of the score. It is never treated as zero or filled in. If less than 50% of the total weight has data, no score or regime is shown. Every value has a maximum age (5 days for daily series, 12 for weekly, 45 for monthly). Older values are marked stale and their confidence is lowered.
How is confidence set?
Each factor starts from a fixed confidence that reflects how reliable the relationship has been (for example, inventories are high, the Brent-WTI spread is low). The overall confidence combines coverage (share of weight with data), agreement (share of scored weight pointing the same way as the score) and staleness. Confidence describes the data behind the reading, not the probability of any price outcome.
What do "Why is WTI moving?" and "What changed?" show?
Why is WTI moving? ranks the factors whose readings point the same way as the latest price move (drivers) and the other way (offsets), by weighted score. These are associations in the data, not proven causes. What changed? lists the five largest weighted changes in factor scores since the previous session and flags new data releases.
What does the AI layer do?
After the deterministic engine has run, Claude (Anthropic) receives a compact snapshot of the scores, values, sources and dates, and writes a short explanation. It must repeat the engine's regime, may only cite numbers from the snapshot, must separate facts from interpretation and may not forecast prices or give trade instructions. Each answer is validated against the snapshot; if it fails twice, no AI text is shown. Headlines, when used, are treated as untrusted third-party text. The AI never changes the score.
What are the limitations?
- Weights and scales are judgement calls, fixed in advance and published here. They are not fitted to predict returns.
- Factor relationships with price change over time; a factor can be right about fundamentals and still coincide with a falling price.
- Weekly and monthly data lag the market. The dashboard shows the period each value describes.
- Some series (futures curve, equities) need a data licence before they can be restated. Until one is in place, they read Data unavailable in live mode.
Where does the data come from, and on what terms?
- EIA (U.S. Energy Information Administration): WTI and Brent spot prices, weekly crude stocks, Cushing, SPR, production and refinery utilization, via the EIA Open Data API. EIA data is in the public domain (EIA copyrights and reuse). Source: U.S. Energy Information Administration (Oct 2026).
- Federal Reserve: the nominal broad dollar index (H.10) and the 10-year Treasury yield (H.15). Dollar index and Treasury yield data: Board of Governors of the Federal Reserve System, retrieved from FRED, Federal Reserve Bank of St. Louis.
- Entered monthly by the Marlin desk: OPEC+ crude production (OPEC Monthly Oil Market Report, secondary sources), China crude imports year on year (General Administration of Customs of China) and the Geopolitical Risk index (Caldara & Iacoviello). Each value shows its publication and month; a missing month reads "Data unavailable".
- This product uses the FRED® API but is not endorsed or certified by the Federal Reserve Bank of St. Louis.
- Neither EIA nor the Federal Reserve endorses Marlin Intelligence or its analysis.
- Futures prices and equity indexes are licensed market data. They are not restated here without a licence and show as "Data unavailable".
See also the disclaimer and Marlin's general editorial methodology.