I've been trying to figure out why the average MPG display (Ave: ...) doesn't match the MPG I calculate using the odometer and the number of gallons I use at the pump. I've read the related posts here:

- The average is not just over the last 15 minutes
- The average is reset with the Reset button
- The Ford folks say it is a running average over the last 2000 data points
- We assume the 2000 data points are the per minute data points we see on the 15 minute graph

What I can't explain is that the average displayed is significantly lower than the average I calculate at the pump. But I have a theory. I think the average display is calculated based on the last 2000 data points, but each data point is the MPG calculated (and shown in the graph). The data points are not gallons used and distance traveled, they are the MPG calculated for each of the last 2000 minutes. So the running average is an average of the last 2000 per-minute averages. This means that the running average is over time, not distance.

If the running average is over time, then a low MPG while going slower (uphill for example) will skew the average lower than true MPG. I think this is what is happening in my case. We drive uphill 8 miles on a bumpy dirt road to our house, getting low MPG.

I tested this theory by resetting the average before driving home. The drive home is in two segments, the first segment driving 70 MPH on the highway and the second segment driving more slowly up the dirt road. At the end of the first segment, the average reads 31.4 MPG. After the second segment it reads 26.7 MPG. The question is: How many MPG did I get during the second segment alone?

MPG Miles Minutes
31.4 42 36
X 8 20

Ending average: 26.7

If the running average data points are gallons and miles, then X must be 2.0. This isn't possible.

If the running average data points are per-minute averages, then X must be 18.4. This is about right.

If I'm right, to get true MPG the gallons at the pump and the odometer reading should always be used. The average display is not true MPG.

Am I right?

--

Here are the calculations I used to get 2.1 and 18.4:

26.7 = ((31.4 * 42) + (X * 8)) / 50
1335 = 1318.8 + (X * 8)
16.2 = X * 8
2.0 = X

26.7 = ((31.4 * 36) + (X * 20)) / 56
1495.2 = 1130.4 + (X * 20)
364.8 = X * 20
18.24 = X

 

I don't know if you are right or not, only that this is one of the most discussed topics ever. All we know for sure is the electronic display is not good for very short segments, but pretty good for tank averages. Mine ( and most ) tend to read 1-2 MPG higher than actual.

Also, your gas tank holds a "nominal" 15 gallons. However, there is 3 gallons of "air space" in there. So, you can actually pump in much more than 15.0 gallons. No matter how hard you try, you cannot replace exactly how much you burnt on the last tank. For this reason, most people trust the gauge for single tank approximations. The BEST idea is to track multiple tanks. GreenHybird is a great place to do that.

I KNOW my car is getting 37.4 MPG overall.
I think it got about 33.1 mpg last tank.
Does that make sense?
-John

 

I don't agree with that last assumption. The reason. I can drive and create a display where the green points on the 15 minute graph are all is significantly HIGHER or LOWER than the average reported in with the white line. I have a picture of that case posted somewhere on this site. The only way that can happen is if the 2000 data points includes points that are not on the 15 minute graph that bring the average up or down. there is obviously no requirement that they can't work with more points than they show in the graph.

My guess is that the average represents the average of the last 33-1/3 hrs driving. Why 33-1/3? If you have 2000 points taken 1 per minute you have 2000 minutes and 33-1/3 hrs.

Also I'd like to point out they picked 2000 points for some reason of their own.... There is no need for a 2000 point limitation. Its not a matter of having enough "memory". You don't need to store all 2000 points to create an average. In order to create an MPG average all you have to do is create a running total of miles traveled (one storage location) and gallons consumed (another storage location). You only NEED two storage locations to find the average MPG. If its 2000 points it was chosen for some specific reason. Probably as a running average so it would give you an idea of the trend of your RECENT driving.

 

Yes.

You're right -- because a fill is not always the same, measuring multiple tanks (with the pump and the odometer) by adding up the gallons and the miles, and then dividing, is the most accurate. It may be the only measurement that is always correct.

For me at least, I can't trust the NAV average at all. Perhaps this is because of my daily uphill 8 miles on the dirt road -- I'm driving slowly but getting low MPG. Perhaps for others, the NAV display is off in the other direction because they normally don't get low MGP while going slow. Usually slow driving is city driving which is high MPG. If I'm right then this explains why for most folks the NAV average is a little higher than true MPG.

If I'm right about the way the NAV average works, the NAV average will be the same as true MGP if you drive the same speed for most of the tank.

Mark

 

Yes, I agree. I didn't mean to imply that the data points are limited to the 15 minute graph. I only meant that they are the same information -- averages over one minutes. If there are 2000 data points, each data point appeared on the graph (at one time) as a per-minute average.

Yes, I understand. But apparently they are keeping a recent running average with some number of data points, since over time the average reflects more recent data.

 

If they use one minute data points, the average will be skewed. For example:

I drive for one minute at 5mph uphill and get just 10 mpg. I drive for another minute at 70mph on teh freeway and get 30 mpg. Average the two data points and you'll get (10+30)/2 = 20mpg. But in practice I used way more gas in the second minute and it should be represented proportionately.

1st minute: 1 min @ 5mph = 0.0833 miles. At 10mpg this = .0083 gallons.
2nd minute: 1 min @ 70mph = 1.167 miles. At 30mpg this = .0389 gallons.
Total miles = .0833 + 1.167 = 1.25.
Total gallons = .0083 + .0389 = 0.0472.
Real mpg = 1.25 / 0.0472 = 26.48mpg

I'd hope that the average is generated from the real-time data acquistion on fuel consupmtiona and milage rather than the snapshot

 

I had hoped this also. But the results of my experiment seem to indicate the opposite.

If you try something like you outlined, but driving more than one minute at each speed of course, that will tell you whether the computer is doing what you hope it is or not. Be sure to reset before starting the experiment.

Mark

 

Yes... if they did the most accurate thing.... problem is it sounds like they did not.

 

But...it can't be using the one minute measures. Here's a much simpler proof: - simply hit 'reset' and watch while you drive. In the first few miles the average is updating constantly and responding to your real-time behavior. If that's how it calculates at the start, it must do the same throughout the sampling period, but as the period becomes longer each piece of real time driving becomes statistically less significant and so the average display appears to be more stable. QED

 

Yes. What I tried to say is that it uses the one minute measures but it stores 2000 of them, not just 15.

Mark