Originally Posted By: SecondMonkey
No. In order to accurately measure down to a few MPG you need a very expensive dyno and some very smart people to run it. "I drive the same route every day and got 1mpg better" just doesn't cut it.
There are simply too many other variables, besides the oil, every single time you get into your car.
What if I told you I was one of those smart people?
In all seriousness - I'm a test engineer and a major part of my work deals with measurement uncertainty.
What you say is partly true and partly false.
First, you've got to look at how repeatable the phenomena you are trying to measure is. If the phenomena isn't repeatable, it makes no difference how accurate or precise your measurements are, they won't make sense. You hit the nail on the head here. If you only compared one tank to another tank, you can't make a good judgment.
However, if you change your time reference and average many measurement samples, your measurement will converge onto the true answer. The earlier poster with the spreadsheet showing a clear trend in noisy data hits the nail on the head. It's statistical fact. Many real-world measurements are quite noisy and it's common to average many, many samples over a period of time to get to the "truth." Thermocouples are good examples of this - quite noisy, yet if averaged over time, they can be quite accurate.
I can show statistically that I can indeed drive the same way, day to day. Over many tanks of gas (about 18,000 miles worth), when I'm driving the truck to and from work, my fuel mileage hangs out at about 18.8 to 19.2 MPG. If you think about this, it's a big heavy truck that's as aerodynamic as a brick, I'm a middle-aged guy who doesn't hot rod in it, doesn't speed, and drives mostly in places where speed limits are rigidly enforced and speeding penalties are quite high, so it make sense.
Now, if I'm driving the Mustang GT, all bets are off as I'm a very moody driver in it. One tank it's 16 MPG as I'm beating the [censored] out of it and the next may be 24 MPG as I'm driving like my inner Grandpa, enjoying the ride with the top down.
With a measurement that is inherently repeatable and not noisy, you need not gather as many samples to accurately show a trend. I have no problem detecting a 0.5 MPG increase, as that's clearly statistically different from my demonstrated consistent base reading and outside the known "noise" in my measurement system.
With this repeatability, I can also tell when I've got a bad data point. For instance, if I compare four tanks and get 19.0, 19.0, 19.2, and the last tank is 20.5, I know I've got a bad point... Same with a negative impact, which is why I cautioned my negative report on non-TCW3 oil as I ***may*** have a bad point.
There is no way without more and longer term data that I could tell you that with more than 1/2 MPG precision. If I had many more samples, I could, but I don't. Four tanks of repeatable fuel mileage is repeatable fuel mileage.
Accuracy is another thing, but thankfully, we're not too concerned with accuracy. Why? These are comparative evaluations. IE: let's say I'm actually driving 50 miles when my odometer only shows 40 miles. As long as that error is consistent, it drops out in the math. (I'm under-reporting MPG in both cases equally.) I can make the comparison.
Precision is the remaining item. If the gas pump only read in single gallons, that's not a precise measurement. If the pump reads 20 gallons dispensed, there could be anywhere from 19.5 to 20.4 gallons dispensed in reality. Thankfully, precision on today's gas pumps are quite good. Same with odometers.
So, bottom line, you can't make a sweeping statement unless you've carefully looked at what you're measuring...
thanks,
ben
