[jtolmar]

It works if $25 of gas lasts long enough to get to a different price (harder at a bad price), and your tank has sufficient room to hold $25 of gas (harder at a good price). Those constraints make it worse than real dollar cost averaging, but it should still work.

You can do better if you can predict gas prices but that's sort of the point of DCA - it's a strategy that works without insight.

Edit: this strategy has a higher sample rate when gas prices are bad. This causes a problem if prices are temporally correlated, which they are, unless the sample distance is large enough that the price has enough time to become completely randomized, which is a big ask.

[lucas_membrane]

> a strategy that works without insight

Doesn't work if gas price changes are log-normally distributed.