Open methodology · Current regional model

How GasPal helps you decide when to fill up

Gas prices don't move randomly. GasPal watches the same kinds of market developments that influence gasoline wholesalers and gas stations, then offers a simple, regional recommendation: fill today, consider waiting, or mixed signals.

Built for driversU.S. gasoline marketsNo claimed accuracy percentageNo exact next-day pump-price prediction

Built from people who price gasoline

Why we chose these indicators

GasPal's founder spoke directly with 20 gas station owners and 20 wholesale gasoline suppliers to understand how gasoline is actually priced. The work built on earlier local gas-price research in Gainesville, Florida, plus experiments with two previous timing models.

What we learned

Wholesalers use relevant physical gasoline markets and their latest closing prices when setting wholesale offers. Stations also respond to their acquisition costs, nearby competition, deliveries, and demand. Refinery outages, pipelines, fuel availability, and unexpected news can move expectations before pump prices fully respond.

Those conversations and previous model tests motivated a more region-specific, immediate framework rather than relying on one national smoothed trend. Interviews explain the choice of inputs; they do not prove the model predicts correctly. Earlier experimental models were not selected as the best basis for the current public recommendations.

The result you see

Three possible recommendations

Consider filling todayUpward price pressure. Some stations could become more expensive.

Consider waitingDownward price pressure. Prices might ease if you have enough fuel.

Mixed signalsShort-term clues disagree or are insufficient. See the longer trend and cheapest nearby stations.

Signal strength appears as Weak, Moderate, or Strong, depending on supporting versus opposing evidence. It's a description of agreement between indicators—not a probability or verified accuracy rate.

What GasPal watches

Four immediate price drivers

1. Regional wholesale gasoline

Prices wholesalers watch in their physical gasoline market. A new regional closing price is our primary market signal. When the current regional quote hasn't been published, we retain the previous close and use newer national gasoline futures cautiously. Where no representative publicly available daily regional spot quote exists, national futures are only low-weight background: verified local repricing, PADD inventories and credible regional supply news take priority.

2. Gasoline in storage

The U.S. Energy Information Administration reports weekly gasoline stocks for five large regions (PADDs). Lower-than-before stocks can signal tighter supply; higher stocks can suggest easing pressure. This is regional context, not the inventory of your neighborhood station.

3. Supply disruption news

A reported refinery fire, actual outage, fuel-pipeline interruption, fuel-terminal disruption, or a completed restoration of operations can affect regional supply quickly. A credible, dated operational report may influence a recommendation before official confirmation, but remains provisional. Planned refinery purchases or future restarts, investor analysis, legal and political debates, airport terminals, and unrelated natural-gas pipelines are not immediate gasoline-supply evidence and do not earn directional news points. Older articles can provide background, not a fresh signal.

4. Nearby station repricing

GasPal compares the same stations across three complete daily noon snapshots. The most recent ~24 hours (latest two snapshots) are the primary directional signal: a station counts as raised (red) or lowered (green) only if its reported pump price moved by at least 5¢ per gallon with a newer provider timestamp during that latest interval. The preceding ~24 hours are kept only as context to see whether changes continue, reverse, or have stopped. Old movements do not vote for today's direction. Three snapshots are required to verify the context, not to give older prices equal weight.

What each time period means: A gasoline spot or national futures change is one trading session, closing value versus the preceding trading close, not a week of movement. EIA publishes regional wholesale spot closes. National futures use Yahoo/MarketWatch reported daily closing values, not independently confirmed CME settlement; a rolling futures contract can change underlying contract month and may distort a close-to-close difference. Because the provider does not supply a verified same-contract identifier, GasPal now excludes continuous national futures daily changes and rolling national background trends near month-end contract rollover from directional scoring. Displayed values remain historical context, not evidence of a forward move. Gasoline inventory changes compare two weekly reporting periods, with a publication date separate from the week-ending date. Supply news uses article publication timestamps (up to 72 hours for scoring; older reports displayed for context). Station repricing scores the latest approximately 24-hour noon-to-noon change. The earlier daily interval supplies context, not a second directional vote; station timestamps identify when the source last reported a price, not necessarily the precise moment the station changed it. Longer national trends explicitly say 10 completed trading sessions.

Why the latest day leads

Imagine 15 matched stations. Yesterday 12 lowered prices; in the most recent daily interval, none lowered and none raised. GasPal assigns zero fresh local repricing direction, even though yesterday was strongly downward. If the latest interval instead has 4 stations raising and 1 lowering, the upward activity leads the local repricing assessment. The older day is context for whether that represents a reversal; it cannot outvote the new observations. Sample checks, independent confirmation, and other market inputs still determine the final fueling recommendation.

How to read “11 lowered · 1 raised”

Those figures now describe only the most recent approximately 24 hours at matched stations. They are not cumulative 48-hour counts. Yesterday's figures appear separately under Details & sources. For example, a station moving $5.60 → $5.45 → $5.52 is counted as raised in the latest interval, despite still being below the first day's price. It can contribute to a verified upward local trend if enough independent stations agree; we also identify it as a reversal, but do not give that same change an additional vote. A watch label alone does not create extra scoring points.

Outside the 12 scheduled research markets, GasPal can compare short-lived, anonymized repeat user lookups (up to 72 hours apart). Those preliminary comparisons remain unscored; they must not be presented as verified last-day movements.

Data integrity and national coverage: Confirmed local scoring requires three complete daily research snapshots, stable station identities, compatible names/coordinates, recent source timestamps, no flagged anomalies, and at least 6 eligible stations matched across all three snapshots, representing 60% or more of the eligible cohort. Smaller samples carry less weight. The scheduled research network currently covers 12 U.S. markets. Elsewhere in the U.S., GasPal can reuse the regular-gas station results already retrieved when a driver visits to build a shared, short-lived six-day price-observation history. A return visit tomorrow can reveal preliminary price changes without an extra HERE lookup, but those client-observed changes do not independently affect timing-score confidence. Device-supplied station counts do not affect fuel-timing scores. Repeated views of one source quote are never treated as independent confirmations. For more information, see the privacy policy.

Exactly how scoring works

The points behind a recommendation

Each input contributes positive points if it suggests higher pump prices and negative points if it suggests lower prices. The rules emphasize the most relevant recent regional developments, while still allowing high-impact new information to change the conclusion.

See the actual point values and ranking

Indicator	Points	When it counts
Fresh regional wholesale gasoline close	±1 / ±2	At least 2¢, but under 5¢ = 1 point; 5¢ or more = 2 points. Moves below 2¢ are treated as noise.
Most recent older regional close	Up to ±0.9 / ±1.65 initially	An earlier ≥2¢ regional move can retain weight for up to 7 days, halving about every 96 hours. A timestamp-verified local cohort that has moved in the same direction reduces the remaining pass-through weight only insofar as the observed movement has not continued into the latest interval.
Newer national RBOB gasoline futures	Regional-close bridge: ±0.45 / ±1.25; no physical benchmark: ≤±0.35	Where a locally relevant publicly available regional spot price exists but is awaiting a new close, newer national futures can be a cautiously weighted bridge. Where no representative publicly available regional physical benchmark exists (for example, Chicago/Midwest), national futures count for at most ±0.35 points, not as the leading regional predictor. In an unverified month-end futures-roll window they count for zero points. Regional PADD inventory, genuinely verified nearby repricing and relevant dated local supply news keep their full weights. PADD inventory areas are not themselves wholesale-price benchmarks.
Nearby station repricing	Adaptive confirmation: ±0.5 / ±1 / ±2; forward score adjusted	Direction and confirmation use the newest noon-to-noon interval exclusively. Verified physical matches require three complete noon snapshots and a ≥5¢/gal move supported by a newer source timestamp. For 6–7 matched stations, at least 3 moves and ≥40% same-direction share with strong agreement yield at most ±0.5 points; 8–14 require ≥3, ≥25%, and ≥3:1 agreement for at most ±1; for 15+ stations, an emerging move needs ≥2, ≥5%, and ≥2:1 agreement (±1), while corroborated movement needs ≥4, ≥20%, and ≥3:1 agreement (±2). The preceding daily interval cannot independently vote. It provides continuation/reversal context and helps cautiously adjust remaining pass-through: forward score = latest confirmation × max(0.20, 1 − 0.85 × same-direction non-reversed cumulative 48h share × (1 − latest-day share of those cumulative movers)). Reversals are counted in today's direction but excluded from the older cumulative wave. This is a heuristic, not a calibrated price forecast or a measured fraction of market repricing completed. Under 6 matches or inadequate coverage, the local score is zero.
Regional gasoline inventories	Variable, up to about ±1.05 normally	Based on the weekly stock change relative to the region's prior stock level. The initial release can get a 40% boost (up to about ±1.47) before a newer wholesale close. Influence halves every 4 days and stops after 14 days.
New supply disruption / recovery report	Up to ±2 initially	More significant region-specific reports carry 2 points, smaller or broad-region reports 1 point. Influence fades after 24 and 48 hours; reports older than 72 hours are retained as context, not a fresh directional vote.
WTI crude oil, when gasoline pricing is unavailable	Up to ±0.30	Only a last-resort upstream clue. WTI crude is not a regional wholesale gasoline or station rack price.
Observed local 7-day retail trend	Up to ±0.35 context; ±0.15 if verified repricing is already directional	A regional seven-day average must move by ≥5¢/gal, have at least 6 dated observations and 60% daily-direction agreement, and still be moving the same way in at least two of the last three consecutive daily intervals, including the most recent interval. Then it makes a modest contribution to the same total as current wholesale, local repricing, inventories and supply news. Its weight is reduced when verified local repricing has already supplied directional evidence, to limit overlapping station-price information. It cannot independently reach the ±0.85 recommendation threshold, cannot override fresh major operational evidence, and is zero if the latest daily trend has reversed or cannot be verified. These weights are cautious heuristics, not calibrated coefficients.
National 10-session RBOB trend	Weak fallback only	Used only without meaningful fresh directional evidence or a verified ongoing local retail trend. Requires at least a 6¢ sustained move and substantial agreement across trading sessions; excluded during unverified futures contract-roll windows. Never a substitute for local station behavior or regional wholesale evidence. Seasonal fuel-blend context adds zero points.

These are transparent heuristic points, not trained regression coefficients. All cents refer to per-gallon price changes, not projected consumer savings.

How the total becomes Weak, Moderate, Strong, or Mixed signals

Total ≥ +0.85: consider filling today. Total ≤ −0.85: consider waiting. Scores between them produce a Mixed signals outlook unless a permitted weak national background fallback applies. The local 7-day trend has a small capped contribution to the total only when the most recent days still support it; it never independently crosses the recommendation threshold.
Strong: roughly 3.4+ forward-looking points in one direction, with corroborating local repricing and fresh regional wholesale or substantial supply-news evidence, and no important opposing evidence. A largely completed observed wave alone does not warrant a Strong recommendation.
Moderate: meaningful current local or regional support, though not all indicators agree. A genuinely new urgent event may justify a Moderate provisional call.
Weak: limited, older, indirect, or primarily national evidence. National futures alone cannot create Strong local confidence.
Mixed: strong opposing indicators, two substantial indicators on either side, or insufficient directional evidence. Sometimes a sustained longer-term trend still offers context.

A new major reported supply event can override the older closing-price direction if it is later than the close and local station evidence does not clearly contradict it. This is an early-warning mechanism, not a statement that the event is confirmed.

Illustrative example, not a live recommendation

For illustration outside an unverified futures-roll window: national gasoline futures −1.25 · earlier regional spot close +0.60 · verified local decreases affecting 6 of 27 matched stations −1.62 (confirmation −2, adjusted for 22% already observed moving) · gasoline inventory increase −0.16. Total approximately −2.43 points. This favors waiting, but opposing regional wholesale evidence and uncertainty about which stations have already updated remain important.

Dates and data gaps matter

What if a market closes, or data are missing?

Friday evenings and weekends

The last completed trading-day close—normally Friday—remains the wholesale baseline until the next completed session. New supply news and observed station repricing can still influence the outlook over the weekend.

Today's regional price not published yet

Show the latest actual regional close with its date, retain part of its influence, and consider a newer U.S. gasoline-futures close as a clearly labeled substitute. Regional and national movements can disagree.

No regional wholesale data

Use available national RBOB gasoline futures as weaker context; WTI crude oil only as a last resort. Do not pretend these are the price your station pays.

No verified recent repricing cohort

Require three complete noon research snapshots over approximately 48 hours, with a minimum of 6 reliably matched non-membership stations (and stronger requirements for higher scores) within roughly 5 miles. Incomplete collections, source timestamps that disagree, identity conflicts, ambiguous reversals or insufficient sample coverage cause local repricing to contribute zero points. Elsewhere, a new visit can start shared price history, and subsequent visits can show provisional reported changes; these unverified observations contribute zero scoring points. Nearby reported prices remain available.

Inventory or news unavailable

Label the missing input instead of interpreting it as no supply pressure or no disruption. News scans can miss events, and some headlines may turn out to be wrong.

No immediate timing consensus

Show the actual local 7-day retail trend whenever reliable observations exist. When it remains consistent through the most recent daily interval, add a cautious ±0.35 point context contribution, or only ±0.15 if verified recent station repricing has already contributed a directional vote. A sustained past trend on its own never triggers a fill/wait recommendation. If the trend has reversed or its recent continuation cannot be established, display it as history with zero directional weight. Only when fresh input and ongoing local trend evidence cannot decide does a sustained national RBOB trend offer weaker background context outside unverified contract-roll windows. A seasonal transition never guarantees a decline.

What we do and don't claim

Transparent limitations

GasPal does not currently claim any particular forecasting accuracy, statistical confidence level, expected savings, or exact next-day pump price. The current regional method is an experimental, research-informed directional proxy based on how actual gasoline markets and retailers operate.

Not all stations change prices at the same time. Even complete three-day snapshots cannot reveal the exact start, completion fraction or ultimate direction of a market-wide repricing event: some stations may have moved before the baseline or may move again. The observed movement share is a bounded cohort proxy used to conservatively reduce forward points, not a verified probability of future price changes or guaranteed savings. Market feeds can still contain errors, and even Strong signals can be wrong.

Our intention is to help everyday drivers make more informed, practical fueling decisions anywhere in the U.S. The quality and specificity of underlying evidence varies by location. When the evidence is inconclusive, GasPal still helps drivers compare available nearby gas prices.

Safety first: never drive on dangerously low fuel or take an unnecessary detour solely because of a timing recommendation.

Sources: EIA wholesale market prices, EIA weekly petroleum and PADD stocks, dated public supply-news reporting, completed RBOB gasoline futures, and GasPal's matched station-price research. Model weights are GasPal's own judgment rules; they are not official EIA scores.