I'm aware of the change as my post on page two indicates.
The point is that the reasonable range of possibilities is pretty wide. Extreme examples within that range should not be unexpected from time to time on a forum with tens of thousands of readers.
You know, bell curve stuff. Something I know you appreciate.
Well, if you want to be exact; data that falls outside of a developed curve; that being an extreme example, is usually ignored due to the unreliability of such a data point. Extreme examples tend to be the result of some amount of spurious data (unknown/unrelated/1 off occurance) and since they impact the reliability of a prediction, they are thus eliminated.
For example, the reason they modified the method for MPG calculation is due to the extreme unreliability in the model. A vehicle tested at static, say on a dyno, absent any real life characteristics is a wholly worthless determinate.
I have no doubt that there are a possible occurance of events that could lead the occasional mpg above the determined curve, but I also consider the variables at play to be so numerous and so incalculable as to render that example not predictable.
In short, driving down a steep hill, in N for 4 miles may very well give someone a MPG rating far exceeding 18 mpg.... that doesn't make it a useful example.
This isn't to say 22 mpg on the v8 isn't possible, in particular the 2005 and above VVTi which were able to gain a minor MPG gain; I just haven't seen much data to support the realistic use of that. Extremes are extremes for a reason. Extremes are not useful because they pull the wrong conclusion.
I do not personally consider extreme examples that fall outside of the curve (which 22 is, well outside of) to be accurate representations of data. That is the entire point of a curve and understanding standard deviation and a confidence interval.
This doesn't mean impossible, as it is rare to say "impossible". Instead it means highly improbable and not supported by the data. It means that all those little things that destroy data reliability LIKE:
Spurious data (Illegitimate, incorrect)
Unknown Variables (Variables at play unaccounted for)
Confirmation Bias (A BIGGIE. The desire of the person to make something occur)
There are more, but these are the biggies. When you move far outside the confidence interval, those results are HIGHLY suspect and usually the result of error (whether intentional or otherwise).
This doesn't mean you can't pull some insane mythbusters hypermiling and manage extreme results. It just means that in terms of practical mileage, it is not a legit result.
The new MPG ratings ARE far more accurate than the previous. 22mpg isn't impossible, you are right; I won't say impossible. I just have to say, highly improbable and I am more likely to suspect something else at play.
It is meant to goad people into wars because it is all personal preference.
