"Unglaublich!" It's one of the few words of German I picked up as a kid. I learned from the X-Men. Every once in a while, Nightcrawler [the blue one who's not Rebecca Romijn] would shout it out. It means "Unbelievable!"
And it comes in handy sometimes, especially when I come across a whack-headed scheme as whack as the one offered by the, um, Institute for Empirical-Statistical Diagnostics in Rodgau, Germany. For just 24 euros--there's a half-price sale in May, only EUR11.90!!--these German crackpots will use their secret statistical analysis method to predict the gender of your unborn child "in the first weeks" of pregnancy.
All they need is the date and length of the mother's menstrual cycle and the birthdates of both parents. That's all! And their method is 100%
accurate guaranteed, or your money back. Unglaublich, you say? How EVER does it work?
Let's look at it empirically:
You pay them EUR25.
They flip a coin to predict the sex of your child.
Probability that they're right: roughly 50%.
Amount of the fee they keep when they're right: EUR23.90
Probability that they're wrong: 50%
Amount of the refund owed if they're wrong: EUR23.90
Estimated percentage of people who will actually seek their refund by mailing in a birth certificate after eight-plus months: 10%*
Worst-case scenario: they refund 100% of wrong predictions.
Expected profit per coin toss: 0.5 * 23.90 = 11.45
Most-likely scenario: they refund 10% of wrong predictions.
Expected profit per coin toss: (0.5 * EUR23.90)+(0.5 * (0.9 * 23.90)) = 22.71
Someone should double check, because I think this joint's real name is the Institute for Empirical-Statistical Ripping Off Total Knuckleheads. Unglaublich.
* Based on the Promotion Marketing Association's 2005 study on mail-in rebate response rates, which can range from 2.7% overall to 30% for narrow, motivated groups. All response rates are negatively correllated with the length of time of the offer, though. No sane rebate takes nine months, so I took the lowest response rate and quadrupled it, just for good measure.