GasPredictor.com
Home

Select Local Prediction:





Mobile Version
  • See tomorrow's gas prices on your phone or tablet.


  • How to Use Gas Predictor
  • Exactly what do these predictions mean?

  • Track Record
  • Over 99% correct since November, 2008

  • Local Gas Prices
  • Current prices in various U.S. cities.

  • Gas Saving Tips
  • Information you can use.
  • Gas Saving Articles.
  • Gas Saving Software.

  • About Gas Predictor
  • Privacy Policy
  • We don't sell or otherwise abuse your information.

  • Contact Us


  • GasPredictor.com's Track Record Summary for Prior Years

    Our published forecasts for 2008 through 2015. Success rate for this period: 99.50%.

    More Track Records

    Summary, 2008 through 2015

    Year First Prediction No.
    Predictions
    No.
    Incorrect
    % Correct Highest
    Upward
    Pressure
    Highest
    Downward
    Pressure
    Highest Price Lowest Price
    2008 11/07/2008 37 0 100.00% Unknown Unknown Unknown Unknown
    2009 01/02/2009 252 1 99.60% 0.587 -0.330 2.632 1.858
    2010 01/04/2010 252 1 99.60% 0.281 -0.348 3.024 2.522
    2011 01/03/2011 252 3 98.81% 0.389 -0.497 3.938 3.029
    2012 01/03/2012 252 2 99.21% 0.337 -0.294 3.954 3.242
    2013 01/02/2013 252 1 99.60% 0.325 -0.310 3.781 3.244
    2014 01/02/2014 252 0 100.00% 0.239 -0.380 3.767 2.258
    2015 01/02/2015 252 1 99.60% 0.321 -0.336 2.862 2.016
    Summary 11/07/2008 1801 9 99.50% 0.587 -0.497 3.954 1.858

    Notes

    • Raw data for the year 2008 were lost in a computer accident. We still have our published newsletters for that year, so we know how many predictions we published and how many were incorrect, but we do not have the exact prices or pressures we calculated.
    • Prices shown are the highest and lowest of the average of our "second lowest" prices, averaged among the twelve cities in our National Prediction Model.
    • Values shown for "pressure," a key indicator of which way retail prices were going to move the next business day, are as they were calculated at the time. Changes and adjustments to our algorithm since then mean that our current algorithm would likely calculate different pressure for the same circumstances. Notice that pressures are somewhat less extreme in more recent years.
    • Even though our success rate for 2011 was less than 99%, our cumulative success rate never fell below 99%, so we stick to our claim: Over 99% correct since we began publication.

    Incorrect predictions and modifications to our algorithm

    • First incorrect prediction 8/14/2009. Average fell 1.4 cents per gallon when we predicted weakly upward. We adjusted some factors in our algorithm, but did not make substantial changes.
    • Second incorrect prediction 5/12/2010. This resulted from some cities in our 12-city model having downward pressure while others had upward pressure, and the average retail price moved the opposite direction from what the average pressure suggested. We modified our algorithm to inlcude a range of both upward and downward movements in the average when pressures among the twelve cities are mixed. This is when we began publishing the predicted range of average prices when pressures are mixed. This is the most substantial change we made to the algorithm since we began publishing.
    • Third incorrect prediction 3/16/2011. This resulted from near-term upward pressure and mid-term downward pressure at the same time. We modified our algorithm to account for this, the second-most-substantial change since we began publishing. We would soon find out that it wasn't enough.
    • Fourth incorrect prediction 8/5/2011. Our local predictions were also incorrect in six cities, the first time we were incorrect in more than one city on the same day. This was another instance of up/down pressures, but more extreme than the one in March. We adjusted some factors in the algorithm, but this was not a substantial change.
    • Fifth incorrect prediction 8/26/2011. This was the weekend of Hurricane Irene, whose impact on gasoline distribution (and prices) reached across the whole country, much wider than we expected. Our local predictions were also incorrect in 15 cities, the second time we were incorrect in more than one city on the same day. It was not a good month. We adjusted some factors in those parts of the algorithm that deal with natural disasters, but this was not a substantial change.
    • Sixth incorrect prediction 8/2/2012. Our local predictions were also incorrect in seven cities. This was a lingering effect of a pipeline outage. Prices had already adjusted for the outage, but when it was announced that repairs would be delayed, retail prices shot up faster than they had in the immediate aftermath of the outage. We did not make any adjustments, thinking this was a singular anomaly. We would soon learn.
    • Seventh incorrect prediction 8/13/2012. This was yet another double-whammy from a supply disruption (a refinery fire this time) where retail prices rose more quickly in response to an annouced delay in repairs than they had in response to the fire itself. We made some minor tweaks to our algorithm that would account for this double-whammy in cases where a natural disaster or other distirbution problem lasted more than a few days.
    • Eighth incorrect prediction 10/11/2013. Our local predictions were also incorrect in two cities, both of which happened to be among the twelve in our National Prediction Model. One city underwent a dramatic price war that erupted without warning, and the other was affected by regional downward pressures having more effect than we expected. We made minor adjustments to that city's regional-coupling factors, but no change to our oveall algorithm.
    • Ninth incorrect prediction 7/10/2015. Our local predictions were also incorrect in two cities, both of which happened to be among the twelve in our National Prediction Model. This was yet another double-whammy phenomenon that began with a refinery fire in California. Retailers reacted late to the realization that California could not readily import gasoline from surrounding states due to environmental regulations. We did not make any further refinements to our algorithm, as this is a problem specific to California and not likely to be repeated often.
    • Price wars used to be a major factor in our incorrect predictions, but in recent years of very low prices, there simply have not been any real price wars. As we begin to return to the regime of high gas prices, we will be able to see if our price war factors have been adjusted well enough.

    More Track Records

    Our prediction algorithm is designed to predict the change in retail price of regular unleaded gas at the second cheapest gas stations. Each day, based on that prediction, we recommend whether to buy gas today, or whether you would come out ahead by waiting until tomorrow. We check our prediction against the most recent available prices at the second cheapest gas stations in our test areas. If you would have saved money, or not lost money, by following our recommendation, we call that a successful prediction.



    Disclaimer: This Web site, and all of its predictions and prediction devices, are for educational and entertainment purposes only. We will not be responsible for incorrect predictions, or for any damage or losses you may incur as a result of using these predictions. While we believe that our prediction algorithm works, you must accept responsibility for your choice to use this information.




    [ Home ] [ Mobile Version ] [ How to Use Gas Predictor ] [ Gas Predictor.com's Track Record ] [ Local Gas Prices ] [ Gas Saving Tips ] [ About Gas Predictor ] [ Privacy Policy ] [ Contact Gas Predictor ]