If recording more increases your % error then you shouldn't even be watching it to begin with. Your chance of error goes down when the number of samples goes up.
Here is a quick example of why short term averages are bad. For easy math (and since I'm not positive on the official rates other than the data stream rate) we'll say the programmer takes 5 readings per second.
Lets say you have two different calculations going. One set of calculations has a small number of readings at 1000 readings and the other has a large number of readings at 10000 readings. If the current average is 16 mpg for both then it is easy to see how a single stretch of road could impact the mathematical formula.
You climb up a hill with an IECON reading of 5mpg for the 10 seconds it takes to climb up that hill. Then you slowly roll down the other side of the hill with zero throttle input (maxing out the IECON at 40mpg) dragging on the brakes (to not speed and ram the little old lady ahead of you) for 15 seconds. Your average in this section of road is 26 mpg according to the programmer.
Factor that into the "small" calculation and once you are back on the road your average is 17.1. The large equation comes out to 16.1. If you then compare to the "full" amount of 100,000 readings, you end up with 16.0, or no noticeable change. Keep in mind, this is the impact of only 25 seconds of drive time, now consider what happens when you are cruising through traffic or hilly areas where you are switching between the gas and brake pedals as well as changing speed often. If you were able to freely roll down the hill to shorten the recording time, the average would not be thrown off as much. Unfortunately the computer cannot account for elevation changes, lack of throttle input, resistance from wind and tires, etc. Until then the average will not be completely accurate except under ideal conditions.
|