← WatchtowerEnergy prices, futures and production impacts
Energy in Forecast Lab combines source-dated prices, delivery-month contracts, historical-data research projections and production-region evidence. Its scope is global benchmarks plus selected regional markets, not every fuel seller or country.
Price identities and units
EIA daily spot benchmarks are wholesale prices; weekly retail gasoline/diesel includes taxes. EU Weekly Oil Bulletin consumer prices include taxes. EUR per 1,000 litres is explicitly divided by 1,000 for EUR/litre; fuel-oil tonne prices remain EUR/tonne. US gallons are not imperial gallons. No currency conversion is implied.
World Bank monthly benchmark averages include crude oil, coal, regional gas and Japanese LNG. Benchmark definitions differ: Australian coal uses Newcastle futures from February 2022; South African coal is a forward-month reference; LNG Japan is an import-price reference. These are not all spot quotations. Missing source values are never zero-filled.
Germany/Luxembourg electricity uses SMARD hourly day-ahead auction prices. Complete 24-hour UTC days are averaged without load weighting. Incomplete days are omitted. Future delivery-hour prices already published by the auction are shown separately from historical daily means. These are wholesale prices, not household tariffs or Watchtower predictions.
Futures
Sampled WTI, Brent, Henry Hub gas, RBOB gasoline and ULSD/heating-oil contracts have explicit delivery months, native units, quote times, retrieval times and source links. Provider symbol, instrument type, currency and month labels are checked; the Brent provider name omits the month, so the standard NYMEX symbol month/year code is disclosed as its basis. Quotes are delayed references, not official exchange settlements.
The sampled curve starts two calendar months ahead and includes months +2, +3, +4, +6, +9 and +12. No missing months are interpolated. Curves are market prices for specified delivery contracts; storage, financing, liquidity, delivery basis and risk premia mean they are not unbiased expected spot prices or guaranteed investment returns. Quotes older than four days are labelled stale.
Historical-data research projections
A fixed, predeclared model combines persistence with 25% of the preceding 20-observation average drift. Extrapolation of drift is capped at 20 observations. Additive prices permit historically negative oil/electricity observations. Horizons count published daily, weekly or monthly observations, not guaranteed calendar or exchange-session dates.
Evaluation uses up to 600 recent observations, disjoint target windows, and only data preceding each projection origin. At least 20 evaluation windows and 120 observations are required. Mean absolute error is shown beside a persistence benchmark, in the same native price units. Current-vintage reconstruction is not original-vintage or prospective performance, and superiority is not established.
The displayed range adds the historical error 10th and 90th percentiles to the current projection. It is an empirical research range, not a calibrated probability interval. Daily observations older than 14 days, weekly older than 21 days, monthly older than 75 days, source failures, or captures older than two days withhold numerical projections. These are unissued research calculations; no forecast is backdated into the existing Forecast Lab ledger.
News, disasters and production
Watchtower's commodity-production data supplies source-specific country output, year, unit and same-basis world shares where available. Severe/extreme, unexpired public weather-warning messages are joined by country. Country co-occurrence is a review priority, not facility-level polygon overlap, verified damage, shutdown or lost output. Country output must never be multiplied by warning counts. Missing or partial warning coverage does not imply safety.
Recent energy-related publisher reports retain titles, source links and publication times. Claims are not independently confirmed, and overlapping reports are not treated as independent evidence. Current news and hazards are not inserted into old training periods or silently turned into price coefficients. Inventories, spare capacity, trade restrictions, shipping, refinery constraints, demand, FX and taxes remain mechanisms to investigate.
Explicit supply scenarios
A user chooses a reported producer, assumed local production loss and assumed absolute short-run net supply/demand elasticity. World supply loss = same-basis world production share × assumed local loss. Indicative price sensitivity = world supply loss ÷ assumed elasticity. The result is an authored static sensitivity, not a probability or forecast; absent world totals withhold numerical propagation. Default assumptions are illustrative, not fitted estimates.
Power-station measurements
Infrastructure retains WRI Global Power Plant Database 1.3.0 (2021), with 34,936 source records. Nameplate MW/GW is power capacity. Annual MWh/GWh/TWh is energy over the displayed reporting year. Source-model estimates remain separate from reported generation. Reporting dates are historical, not a claim of present-day output or complete global inventory. Negative net generation is retained; missing values remain unknown. No capacity-times-hours calculation is represented as actual generation.
Sources
Method version energy-outlook-v1 · 14 October 2026. All sources remain subject to their own coverage, revisions and update schedules.