Law of Averages

[Big Dave]

I prefer road-testing over dyno testing.
Reasons:
1. If I could afford a lot of chassis dyno time, I would be wealthy enough to not give a rip about fuel economy.
2. Chassis dynos suitable for diesel pickups are fairly rare. Diesels make ton-miles of torque at low RPM and most hot-rod-shop chassis dynos are made for high RPM engines.
3. Dyno testing cannot pickup the advantages derived from aerodynamic mods. Aero load (the biggest single factor in the Road Load Equation) can only be guessed at on a chassis dyno.
4. The whole point of trying to reduce fuel consumption is to do so in the real world.
5. I dislike the EPA and their general approach to doing things.

[Big Dave]

There is a lot to be done to make road testing more consistent.

A. I run the same course every day ? my daily commute ? and by luck it is an almost perfect testing circuit. It is 45 miles door to door. About 16 miles of it is urban/suburban. Lots of stoplights. About 15 miles of it is Interstate highway. The rest is two-lane state routes with fairly heavy but smooth-flowing traffic.
B. I run the same speeds. My truck has a Auxiliary Idle Control (it can run a PTO) that includes a very accuracy digital tach. On the Interstate, I run 1700 RPM (70 MPH) plus or minus 10 RPM. On the state roads, I run 1450 RPM (60 MPH) on the two-lane roads. In town, I run with the traffic. I do not make a habit of running under the posted limit as this is simply not a fair way to test and I really don?t care to get into a gunfight with some road-raged dude. The need to avoid bending the sheetmetal trumps everything else.
C. I run comfortably. That means I use the A/C and heater as needed. My circuit is heavily traveled enough that straddling the crown of the road is out of the question. I coast wherever I can, and try to maintain momentum (I run some rather orangish lights at times) but I don?t draft semis (although I know this works, my old ticker will not take the sustained stress). I idle at lights but shut off when caught at a railway crossing.
D. I fuel up at the same station all the time. I do not run B2 or B20 while testing. They have a lower heating value and that confounds my tests.
E. I fuel up every 500 miles. That is about a week?s driving and is in the neighborhood of 20 gallons every time. Without precision volumetric or gravimetric instrumentation, it is impossible to fuel up to the same mark every time. A minimum test to me is four fill-ups ? roughly 2,000 miles. Being a fairly high-mile driver this allows me plenty of chances.
F. I do not test in winter other than just to keep my driving style sharp. Not only does the increasing air density confound comparison, but the people?s driving habits change in inclement weather and (seeking to avoid wrecks) I conform for safety. Snow or slush also impose and unmeasurable increase in rolling resistance. My test season is May 1-October 10. During season, I include mileage driven in the rain and regardless of wind direction.
G. I generally try to make one change at a time but sometimes project schedules overlap and my test season is short.

In a nutshell, I use a fairly long test to average out variations.

[Freedom_man]

I keep a running log and tally of how much fuel I use and how many miles I drive to keep track of my MPG and I don't look at false highs I look at my over all running logs to see if something is working per design or not.

I need to run about 3 more tanks before a switch spark plugs to see if Halo Plugs work better. I read a post in a miata forum that they gave the same performance as a side gap plug does.

I also just had my carburetor rebuilt so my MPG has dropped. I also need to re-adjust the choke so it stays on about 2-3 minutes longer than it does now.

[Mighty Mira]

I agree that if you are concerned about accurately measuring fuel economy through tank/odometer, large tankfuls are the way to go. The problem with large distances is that differing speeds and number of stops from a particular speed will play havoc with your FE.

Obviously, an instantaneous meter utilizing the injector signal is more accurate, as that's what the car uses to meter its fuel. At least you can get an idea how much fuel the car is using at an instant, although to use that to get other, more relevant figures (drag coefficient, rolling resistance coefficient) is plagued with problems. And changes in wind and temperature will have large effects on engine load to go a constant speed, and hence instantaneous mpg.

Then there is the driver aspect to consider. We are all obsessed with FE here, and over time our driving technique improves. We learn to P&G, then to bump start the car properly, to slow our speed down, etc etc. All these non-mod related improvements will hamper our ability to measure modifications.

Originally Posted by MetroMPG

and i've discovered that ironically, the more efficient your car is, the more difficult it is to get consistent measurements.

And that's even worse, although the whole point. I used to be able to get results within a week, now I have to wait at least 8 days for a whole tank.

[DarbyWalters]

It is nice to know what a run of 100% highway yields...but overall the 90 day average is great. Once you reach the first 90 day mark, you try to improve your 90 days average each fillup.

As for "testing", small samples are always suspect. I used to work for Vehicular Testing and we would make runs of 1,000,000 miles using maybe 8 vehicles running 22 hours a day...but that IS big money. Whether tire testing, oil comparisons, ect...the more data the better...but we are just individuals so small samples can be a bit more accurate if YOU make them that way.