Doesn’t work that simply. The basic rules of statistical analysis require at least 30 data points for each sample to be able to determine if there is a
statistically significant difference between the test platforms. It requires all variables in each test to be the same; that’s what kschachn was inferring. You may have seen 25 and 26.4mpg, but if the standard deviation was +/- 1.5 mpg, you’ve gained nothing.
Air temp, wind direction and speed, atmospheric pressure, road condition… the uncontrollable variables will always induce more noise into your measurements than any actual mileage improvements ever would… proving that you can never attribute the improvement (or detriment) to a single variable when you’re attempting to test in the manner you reference.