Mixed messages

After my fill-up on August 10 (calculated miles per gallon of 23.90 so climbing back up where I had grown to expect) I installed a K&N high performance air filter. I wanted to see if such a product could increase my fuel efficiency. I was amazed to see that the instant mileage indicator, which I consult frequently when driving, seemed to show a distinct and significant increase. For example, on normal, relatively level stretches of freeway at 55 miles per hour I typically see the indicator jump between 31 and 32 miles per gallon. After installing the K&N filter, the same stretches yielded indications of 33 and 34 miles per gallon. Amazing and quite significant. This should be my best tank full yet, huh?



Think again. Though the driving I did for this tank full was quite typical of my average driving regime, the fill-up from yesterday (August 18) yielded 21.16 miles per gallon. This surprised me as the instant mileage indications led me to expect something like 24 miles per gallon early in the tank. As the fuel quantity indicator went down though, it was clear that I wasn't going to see a good number.



I surmise that one of the major contributors to the variance I see in my results (current standard deviation, the square root of variance, in calculated miles per gallon evaluated at fill-up is 1.71) is inablility to fill the tank to a precise level. That is, if my previous fill-up was short of normal and the current one is above, I'd calculate a lower mileage. The opposite is also true. Such effects clearly average out over a period of time and thus I follow my five tank and ten tank moving average to account for this.



But it's hard for me to see how it could be the entire explanation of this particular fill-up. I put 19.104 gallons in the tank after driving 404.2 miles, thus getting 21.16 m.p.g. For me to have gotten 23 m.p.g. I would have to have only needed 17.565 gallons. That's a difference of 1.539 gallons. It could have happened if the earlier fill-up was, say, about 0.77 below the "average" spot and the most recent was 0.77 above that spot. But it seems like that's an awful lot of variation.



Because I'm aware that the fill-up point is so important to the calculated mileage, I fill it absolutely as full as possible because that seems to be the most accurately repeatable amount. Though I know one is not supposed to "top off," I invariably do. I top off until it's only possible to add a couple of cents worth of fuel (say 0.006 gallons) before the automatic shut-off turns it off. I know that if I didn't top off, the variations would average out to an accurate number over time but I'm too impatient to wait for the averaging out process.



But I'm having a hard time reconciling the low mileage calculated for this tank full with the noticeably higher freeway instant miles per gallon readings. I guess I'll give it another tank full or two before I delve into it more deeply, though I don't know what that delving would invlove.