Automobile Mileage Forecasting

In losing my father-in-law, I was fortunate enough to receive his 2009 3-Series BMW.  My first car was a 1957 Chevy 4-door 210, the second was a Chevy Blazer that I drove until President Obama gave me a deal on it, and most recently I was driving a base model 2010 Camry.  As a frugal guy, the E90 is a wonderful vehicle and driving it is fantastic.

The vehicle wasn’t new for me – it had been borrowed in the past for a few road trips.  I’d never paid attention to the mileage.

I’m now five full tanks in of having monitored the gas consumption in combination with its ‘Mileage to Empty’ calculation.  I’d thought I’d noticed in the past that there was a sharp drop-off in expected mileage over time.


Mileage on the y-axis and observation on the x-axis.  Four tanks in an E90 of odometer and ‘Mileage to Empty’ to show forecast fuel efficiency.

The trend started to show here, but I wanted to try and see things more clearly.  I took all four tanks of gas and normalized them, such that the thank was filled up at ‘Mile 0’ and then followed it through the rest of the data points.  I took an initial observation of ‘Total Expected Miles’ which was the odometer reading at Mile 0 + the estimated MTE reading.  At each further point in time (which on the x-axis is the distance traveled since Mile 0), I took the updated ‘Total Expected Miles’ and divided it by the initial Total Expected Miles.

The results are consistent and a bit disappointing.  At the end of the tank, the values dip below 85% of the initial TEM.  For drivers and owners this makes a big dip given it comes late in the life of a gas tank.


% of Total Expected Mileage on the Y-Axis as a function of miles driven for four tanks of gas. After 210 miles the tanks total expectation start to drop to 80% of the initial expectation.

About flybrand1976

Find me on twitter @flybrand.
This entry was posted in Industry and tagged , , , , . Bookmark the permalink.