See also:
http://ecomodder.com/forum/showthrea...ures-2721.html
---
It would be nice to document their methodology (always good to be able to critique - to point out both the flaws & the good points).
So far, here's what I was able to note:
test car: 2004-2007 Ford Taurus
test methodology:
Quote:
- Lay out a driving course that simulates normal city driving with a series of stops, turns, etc.
- Bypass the car’s gas tank and fit it with a fuel cell that can be removed and weighed in order to measure precisely the amount of gas used during each test.
- Make sure tires are inflated to the recommended pressure of 35 psi for the control test.
- Drive the course and calculate the gas mileage.
- Repeat step 3 with tires extremely under-inflated (10 psi) and extremely over-inflated (60 psi).
- Repeat step 3 with tires slightly under-inflated (30 psi — 15% lower than recommended pressure) and slightly over-inflated (40 psi — 15% higher than recommended pressure).
- Analyze the results.
(source: TV Review: Mythbusters 8.23 – “MiniMyth Madness” | Fandomania )
|
So obviously the biggest flaw in their testing was the attempt to duplicate the exact same "real world", variable speed, stop & go type driving for each run. Impossible to do!
Even though it was on a closed course (removes variable of other traffic), and they attempted to time the stops & acceleration to ensure consistency, it's still silly. For best data, they need to remove the driver from the equation. Of course, footage of a car on a dyno would be boring TV (for the average non-Ecomodder viewer
), but would be the most accurate. A TV compromise: Grant should build a programmable "driver robot" for future tests!
On the plus side is their fuel measurement technique - much better than a ScanGauge! (or equivalent)
Could also question the data based on ambient conditions (depending on how long the tests took, changes in temperature as the day progresses could significantly affect results).
Could also wonder about how warmed up the car was before the first run & whether they allowed it to cool significantly at any point between runs (union lunch break between tests??)
And they never repeated the control at the end of the test, so we have no hint of how much inherent variability there is in their numbers.