Exactly right, jgbrowning!
Here's what I did: of all the choices, I let you choose one of them. Then I bundled all the other choices together. If you chose correctly originally, I selected one of the bundled choices at random to represent the bundle; if you chose incorrectly originally, I select the correct number from the bundle to represent the bundle.
99% of the time, you chose incorrectly originally; that means 99% of the time, I'll be choosing the correct number from the bundle to represent it, and 99% of the time, you'll benefit from switching.
1% of the time, you chose correctly originally; that means 1% of the time, I'll be choosing an incorrect number from the bundle to represent it, and 1% of the time, you'll lose by switching.
That's exactly what Monty Haul does in the "Let's Make a Deal" game, except instead of working with 100 choices, he's working with 3 choices.
And that's the explanation that finally convinced me that this statistical trick worked as advertised.
Daniel