[danaris]

There's a classic scam (I'm sure it has a name, but I don't know it) where the scammer gives out predictions for some stock or other: they pick their (large) group of targets, then send half a prediction that it will rise, and the other half that it will fall. After the appropriate time passes, they are proven right to one half, and wrong to the other.

Then they take the half they got right, and do the same thing again. After a few iterations, their pool of marks is smaller...but they also believe that the scammer is never wrong. (I believe the scammer then tells them they need to put a large sum down—which the scammer will, of course, handle for them...and then absconds with it.)

You're focusing on the ones who got it right, ignoring the ones who got it wrong, and calling it smarts. It's very unlikely to be so. The market as a whole should always be treated no differently than the scammer's entire initial pool of marks.

[andsoitis]

John Allen Paulos describes this in his book, Innumeracy.

(Excellent book, by the way)

[xsmasher]

Bob Hope does this in the beginning of "The Lemon Drop Kid," but with horses instead of stocks.