How far off can the Electoral College get?

I have a look at the Current Electoral Vote Predictor 2004 from time to time. I stare at the colorful map, and I look over the numbers a little.

It's pretty well known that our Electoral College could elect a President who did not win the popular vote. I wondered this morning how far from the popular vote a candidate could get and still win the election.

So, I grabbed some population data and came up with some wild premises. First, everyone who can vote, does (because I don't have registration data for North Dakota and Wisconsin and also because it simplifies the math, which I'm nevertheless doing with a computer). Second, there are only two candidates, and everyone votes for one or the other. Finally, everyone votes in the way that maximizes the disparity between the popular vote and the winner of the election.

I compute the number of popular votes per electoral vote, and I sort my states by that metric, which is coincidentally approximately by population. In the 40 least populous states, exactly half-plus-one vote for the winner, and the rest for the loser. In the other larger states, everyone votes for the loser.

As a result, the winner gets 273 electoral votes and 45,314,040 popular votes. The loser gets 160,500,960 popular votes. Thusly, we get a President who got only 22% of the popular vote in an election where everyone who could vote, did.

So, that answers that question. A more realistic model is left as an exercise to the reader.