The Million-Dollar Waitress, from Business Week.
A waitress in Ohio, Mary Sue Williams, may win the million-dollar grand prize in a CNBC stock-picking contest. She was in sixth place when the contest ended on May 25. But a flaw in the way the contest was set up meant that, basically, the five people ahead of her "cheated". They found a way to select stocks to buy at, say, $20 -- and then wait until they went up to $25 before pressing the "buy" button. Because of the way the contest had been programmed, they only had to pay $20 in fake money to buy the stock. (I'm putting "cheated" in quotes because although this is clearly against the spirit of the contest -- you couldn't do this in the real market -- perhaps one could argue that the contest is defined by whatever the computer lets people do.)
Furthermore, it's not really clear how meaningful a contest like this is. From what I can tell, the contest lasted for ten weeks; the winner each week won $10,000 and the grand-prize winner won $1,000,000. Although I can't find information about the prize structure, my guess would be that any other prizes were much smaller. So this encourages risk-taking that nobody would take with actual money. Let's say there's a second-place prize, and it's $100,000. In "real life", most people would be happy with that money. But if you bet -- and I'm using "bet" here because this really does feel like gambling -- all your money on some very risky stock, then you have a shot at multiplying your money by ten. But it's probably a lot more likely that you'll lose it all.
Jim Kraber, on the other hand, played legitimately but at one point had 1600 portfolios. This enabled him to take the sort of risks that someone with a single portfolio -- or even a number of portfolios that could reasonably be played with with hard currency -- would never take, because he could afford to just throw out a portfolio that wasn't succeeding.
Williams admits that "Part of this was luck... a lot of it was a gut feeling, some eenie-meenie-minie-moe, and common sense." Now, I'm not saying that she can't pick stocks -- who knows, she might be quite good at it. But this story reminds me of the following scam. I get a mailing list with 64,000 people on it. (I'm not sure whether this should be postal mail or e-mail; the story, which is not mine and which I learned from a book of John Allen Paulos, predates e-mail.) I tell 32,000 of them they should, say, bet on football team A to win this week, and 32,000 that they should bet on the same team to lose. The team either wins or loses. Next week, I take the 32,000 that got the right answer this week, and split them in half. To 16,000 I say that team B will win this week, and to 16,000 I say that team B will lose. I repeat this six times, until 1,000 people have received six correct predictions. Note that when I begin this scheme I don't know which 1,000 people this will be, but I know they'll exist. Then I try to sell these thousand people my "system". Maybe they'd buy it!