Menu
The adaptive WTI analyst's current forecasting strategy. Load this at the start of every prediction task. This file is generated — edit the state through the mutation tools, not by hand.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
How to use the search_web tool well when grounding a forecast in recent news — phrase cutoff-aware queries, decide what is worth searching for, and weigh sources. Load this before your first search_web call. No scripts.
How to use the search_web tool well when grounding a forecast in recent news — phrase cutoff-aware queries, decide what is worth searching for, and weigh sources. Load this before your first search_web call. No scripts.
How to use the search_web tool well when grounding a forecast in recent news — phrase cutoff-aware queries, decide what is worth searching for, and weigh sources. Load this before your first search_web call. No scripts.
How to use the search_web tool well when grounding a forecast in recent news — phrase cutoff-aware queries, decide what is worth searching for, and weigh sources. Load this before your first search_web call. No scripts.
Reference guide for the repo concierge catalog — domain filters, the search-then-fetch workflow, and bootcamp routing. Load references/catalog-guide.md before your first answer. No scripts.
How to use the code execution sandbox well — parse the JSON payload (not disk files), compute a couple of useful diagnostics before forecasting, and keep the session stateful within a turn. Load this before writing code. No scripts.
| name | wti-strategy-trained |
| description | The adaptive WTI analyst's current forecasting strategy. Load this at the start of every prediction task. This file is generated — edit the state through the mutation tools, not by hand. |
Produce calibrated probabilistic forecasts by combining two evidence streams: statistical analysis of recent price history and web-grounded news context.
At short horizons (5 bd), momentum and recent trend dominate. Trust the trend projection output unless there is a strong near-term catalyst visible in news context (e.g. an imminent OPEC+ meeting or scheduled inventory release).
At medium horizons (10 bd), OPEC+ meeting schedules and US inventory release dates matter. Check for scheduled events in the news context before finalising the forecast.
At long horizons (21 bd), macro demand and geopolitical risk dominate. The statistical signal loses explanatory power at this horizon; weight news context and published analyst consensus more heavily than the trend projection.
Always run statistical analysis (vol-regime, trend-projection) before incorporating news context. The regime classification and trend window directly inform interval calibration.
(No calibration corrections yet. Graduate a confirmed hypothesis to add one.)
| ID | Claim | Confirmations | Refutations |
|---|---|---|---|
| hyp-001 | Point forecasts generated by extrapolating a linear trend are consistently less accurate than a flat-trend (no-change) forecast in the elevated or extreme volatility regimes (annualised vol > 35%), particularly at the 10bd and 21bd horizons. | 0 | 0 |
| Date | Finding | Linked hypothesis |
|---|---|---|
| 2026-06-03 | In the 2025 WTI backtest, the trend-projection model underperformed a simple flat forecast across all regimes and horizons, with the most extreme errors occurring in the elevated volatility regime at the 21-day horizon (Trend MAE of 11.95 USD vs Flat MAE of 3.91 USD, and a negative bias of -3.03 USD). | — |
| 2026-06-03 | A full-year 2025 backtest (247 trading days) of WTI crude oil prices shows that in the elevated volatility regime, the trend projection's MAE was more than double the flat forecast's MAE at 5bd (4.89 vs 2.33) and 10bd (5.28 vs 2.58), and more than triple at 21bd (11.95 vs 3.91), with a strong negative bias (-3.03) at 21bd. | hyp-001 |
| 2026-06-03 | In the 2025 WTI backtest, the trend-projection model underperformed the flat-trend forecast even in the normal volatility regime across all horizons (5bd: 2.60 vs 2.09, 10bd: 3.45 vs 2.83, 21bd: 4.56 vs 3.44). | — |
| Date | Change |
|---|---|
| initial | Strategy initialised with domain priors. No backtest evidence yet. |
