*after the fact*!

Here's a thought experiment. Start with 10,000 stocks. On any given day, for any given stock, the price either move up or down by 1% (49.9% probability each), or move up or down by 25% (with 0.1% probability each.)

Suppose every single day you pick 100 stocks to buy, that you will subsequently sell the next day. You'd expect every day that 0.1 trades would look like insider buys because you'd buy right before they jumped by 25%, and you'd sell the day they jumped. Similarly, You'd expect the same number of insider sells.

In ten years, regulators flagged 18 trades by SAC. In my simple example, you'd have 18 in less than a year.

Wait, you say, SAC's flagged trades happened many days in a row. Okay, fine. Don't buy and sell immediately. Randomize each day, so that there is a chance you randomly accumulate a position over three days, followed by selling for three days to unload the position. This is like the probability of flipping a coin (HHH) followed by (TTT) (or visa versa) at the same time the underlying stock falls in the 0.1% chance of a 20% move up (or down) on day three.

Rare indeed. I bet 18 times in 10 years is not so far off, statistically speaking. (But I'm not smart enough to do that math.)