i would like to point out that the amount of acetone is very small; it will have no effect on the octane rating.

further, by (supposedly) promoting better vaporization, it will result in MORE COMPLETE combustion, and not the reverse as bluegirl's colleague seems to suggest.

further correction: the density of acetone is not the same as water! since the "recommended" dose is around 2.5 oz / 10 gallons of gasoline:

acetone is 0.785 g/mL (less dense than water)
there are 28.3 g/oz
conversion ratio is 36 mL/oz

"recommended dose" of 2.5 oz / 10 gal = 90 mL acetone / 38 L gasoline

my usual full tank ("hindi sagad" is around 41L although i think the tank capacity is around 50L.

i filled up yesterday, full tank shell velocity, added roughly 90 mL of acetone (a 120 mL bottle of 100% is 20 petot at watsons). i didn't use a graduated cylinder, just ratio and proportion (this is not rocket science, folks, it doesn't have to be exact).

also i question the "reagent grade" vs "consumer grade" issue. reagent grade has probably close to zero impurities, but the "consumer grade" is still labeled as 100% -- if it has 5% impurities i seriously doubt that will have any effect.

anyway, i have over TWO YEARS of mileage data for my car. i know one full tank won't be enough for accurate measurement, because in my two years of data, there's a lot of variation.

a more "robust" method which would average out any variances due to driving style, etc. would be to collect lots of data with acetone (at least 10 data points, that's 10 full tanks), then do a chi-square analysis with the old data (i can do this since I have all that old data). the result of the chi-square will tell us if the data with acetone, is significantly different from the old data. simple "i got 10% improvement" is not really conclusive.