Archived timing model experiments
This is the archived track record for earlier experimental models, not an accuracy claim for GasPal's current regional recommendations. See how the current regional model works →
Historical timing beta
Previous model results
Checking results…—Correct calls on meaningful price moves
—Verified next-day station checks
—Recent predictions awaiting follow-up
Loading actual verification records.
Historical five-session model vs. original
We run both models on the same station and starting price. We only score a comparison when both have the same verified follow-up result. A dash means insufficient data, not 0% accuracy.
This comparison is a live 24-hour station test. Historical backtests use different periods and are not interchangeable with these scores.
What do these numbers mean?
- Correct: When GasPal suggests filling or waiting, the price changes at least 2¢ per gallon in the predicted direction.
- Verified: A fresh price from that same station is available at the follow-up check. Repeated visits don't create extra wins.
- Average result: The price difference in cents per gallon, including wrong calls. It estimates an opportunity, not proven savings for an individual customer.
- No clear advantage: GasPal does not guess when the signal is weak. These days are measured separately from fill/wait accuracy.
These results refer to earlier models, not the current regional outlook. No guaranteed accuracy or savings are claimed.
How the previous models worked
The archived beta used a five-session RBOB trend and a cautious waiting rule. The current regional outlook instead includes supply conditions, news and early local repricing.
How the current regional model works →
Original model's historical results
The former daily gasoline-and-oil timing model has a separate historical record. Its different evaluation period cannot prove the new beta's accuracy.
View older model's archived track record →