Law of Averages

[GasSavers_DaX]

This is something that has been sort of bugging me for a while. When performing an experiment, most people are only driving about 50 miles or less [also look in the Gaslog...there are some trips that are less than 40 miles]. This may be giving a false high reading of mileage. Essentially at one extreme, I could fill up, drive one mile, then go back to the gas station and only put in a very small fraction of gasoline. When calculating the mileage, you could possibly get an astronomical number of excellent mileage.

I propose that when doing mileage runs or experiments, we drive as far as possible. I know that once I get my transmission in and after everything I do to significantly increase the mileage of my car, I plan on running out an entire tank of gas on I-285 [Atlanta perimeter]. I know others may not be so lucky as to have a 63 mile oval near where they live, but the more distance we drive, the more accurate the outcome will be.

See http://en.wikipedia.org/wiki/Law_of_averages for more info.

[MetroMPG]

i agree dax.

another experiment everyone should try is to determine your margin of error for this kind of testing: simply do 3 or 4 back-to-back runs & fill ups on the same route without changing anything on the car and then compare your results. i'd be willing to bet that you'll be surprised (i.e. disapointed) at how inconsistent they are. (i was when i tried it on my car.)

and i've discovered that ironically, the more efficient your car is, the more difficult it is to get consistent measurements. the smaller the amount of fuel used, the greater the effect of refueling discrepancies on your results (from thermal expansion, pump variability - if using different pumps - and filling error - e.g. not filling to exactly the same point in the tank). with an efficient car, you have to go proportionately further than a less efficient car to get the same relative accuracy.

anyone who hasn't read fuelsaving.org's position on the inherent problems of on-road testing really should. the whole page is worth reading, but the interesting bit to me is right at the bottom of the page, comparing the margin of error of on-road testing to dynamometer (drive cycle) testing:

The statistical measure of variation, "standard deviation divided by mean", is 12.6% for the on-road data and 2.1% for the drive-cycle data.

[Matt Timion]

agreed. As an out of practice stats nut, things like this are very important.

Another good page from the fuelsaving.org is this:

-snip

Fuel consumption is what most people are interested in. Again the claims are almost always based on users reporting fuel consumption improvements, and not any kind of objective measure. The basic problem here is that fuel consumption is extremely sensitive to driving style, type of journey, even the weather:

Here are some real measurements from my car over a 6 month period. The gridlines are about 10% apart. It's obvious how big the "noise" is, yet people will fit one of these devices and confidently state after just a couple of refills that they have an "11.37% fuel economy improvement", or something like that. Even motoring journalists, who you might hope would know better, are guilty of this. (A cynic would wonder if this is connected to the advertising revenue their magazine receives from the device's makers.)

-snip

read more here:
http://www.fuelsaving.info/debunk.htm

I guess it would be cool to hook up a php chart software that could take your gaslog entries and plot a chart like the one above. It should show the massive margin of error in these types of things.

[diamondlarry]

another experiment everyone should try is to determine your margin of error for this kind of testing: simply do 3 or 4 back-to-back runs & fill ups on the same route without changing anything on the car and then compare your results.

I think I'm the one that you're talking about with the short trips. I have done the 40+ mile trip several times and my mileages are fairly consistant. One day I did two back to back runs(all I had time for that day) and got 50.66 and 50.91. These seem to be fairly consistant. I do agree with you though that longer runs would be the most accurate way to measure efficiency. When you get down to amounts that are less than 1 gallon a tenth of a gallon can make a huge difference.

[rh77]

The only time I've filled-up early was to get a full tank ready for a new test (it was 55 miles). In addition...

For consistency in my own testing, I've changed a few ways in which I fill-up:

* At least half of the tank has to be emptied -- usually around 150 miles or more

* I fill up at the same pump at the same station (luckily it's a Tier 1, and the closest to my house -- QuickTrip by the way, and they're almost always the cheapest)

* We don't have any funky spring-loaded vapor-return pump tips, so I try to insert the filler tube at the same depth, and...

* Let it be done the first time it clicks off (before I was trying to fill it up with a few extra squeezes. (This is really hard for me, because I come from a long line of "fill-er to the brim" pumpers -- and the gas guage isn't buried in the F zone -- it's just shy of it).

The first tank on this procedure was probably an overly optimistic figure, but the second tank brought it back down to reality (which it did -- from 35 to 30). Subsequent tests will probably yield 30-31, if the same conditions permit, which is currently my average.

I sure hope summer yields better mileage, because short of major changes, I may have hit the wall. WAI, Thermostat, radiator baffle, much more efficient driving style, over-inflated tires. I'm almost crazy enough to try Hydrogen, but I'd have to park the car outside (colder temps -- longer open-loop).

BTW, does Acura make an engine block heater for the '98 LS engine? I found one for a 2000 model year... I want to try that next, but they don't have plugs at the airport, so it'll only be effective half the time in the Winter...

What next?

RH77

[Flatland2D]

For anyone that hasn't read my results posted in the tire pressure thread, I think it'd be a good read and example of the consistency obtainable with this method. I'm very confident in my results. If you were to plot MPG vs. ambient temperature and use a best fit curve, each test would be (I believe) within 0.1MPG of the curve. Granted this was under very controlled circumstances unlike normal daily driving. Even so, the ~80 miles I was driving was more than enough to achieve the accuracy I needed.

If you're talking fuel economy for city driving, I think you'll have to use the law of <i>very</i> large numbers, as there are simply too many variables to acheive any kind of consistency no matter which method to test with.

[theclencher]

Originally Posted by MetroMPG

anyone who hasn't read fuelsaving.org's position on the inherent problems of on-road testing really should.

THE LINK SHOULD BE REQUIRED READING TO ALL HERE!

YES I meant to yell!

[cfg83]

theclencher -

Originally Posted by theclencher

THE LINK SHOULD BE REQUIRED READING TO ALL HERE!

YES I meant to yell!

That's why I wish we had "smog check" dudes as members of GasSavers. Instead of going to the bar, we could go to the dynamometer to find out if certain kinds of (non-aero) mods actually work. This would give us the double benefit of measuring emissions in comparison to MPG gains.

CarloSW2

[GasSavers_kevlar]

Hi All, newbie here,
I have the good fortune to drive 45 miles each way to the Steel Mill here in Indiana, and it's pretty flat here. I've been doing all kinds of fuel/acetone/GP7 ratio tests since last spring. I don't do too awfully good as I have a Park Ave S/C, a 3800 V6 Regal, but I CAN get good improvements when the gas is good and the ratios are right....and I have recently acquired a 97 Tercel which I intend to see what MPG's I can muster out of it. As far as the Buicks are concerned, the regal will garner me about 32-34 MPG per tank with a good mix. On a lucky tank, I can get 29 out of the Park Ave.
I am new to the forum, but I am always looking for ways to improve MPG's any way I can! ANY help as to intake and fuel mix would be great. Is there any plusses to having a supercharger? can I modify it for economy? Do K&N filters really help economy?

[Mayhim]

I agree that longer is better. But lets not lose sight of some things...

1) Accuracy vs Anality - Some people aren't satisfied with anything short of a multi-year, multi-thousand dollar effort, while others just want more info than they had. We each draw the line where it's comfortable for us. Sometimes perfect is good enough, and sometimes good enough is perfect.

2) Whole Tank (city/hwy) vs Partial Tank (hwy only) - I think these are separate targets that require different efforts. One needn't do an entire tank on the open road to get an accurate idea of their highway mileage.


If I were attempting to quantify an effort that may only be worth 1/10th mpg, I think it's understood that it would take more than one 40mile trip. On the other hand, if I do two 100 mile trips and see a 10mpg increase I'd say that's a good indicator I'm on the right track. Adding more decimal points will come with time...or a desire to add more decimal points.