Steven Levitt, one of the authors of Freakonomics, discovers poker wisdom: the value of betting hard in the end game of a Hold'em tournament.
So it was 10:15 pm and I figured the latest I could leave for the airport and comfortably make my flight was 10:30 pm. I decided that I would go “all in” with any hand that was decent at all. This would give me a chance to either lose quickly or maybe win quickly. I didn’t announce this strategy to the table, I just adopted it. Over the next 12 hands I was “all in” about 6 times. Four times everyone else folded. I won one of the other two hands, and lost one against someone with fewer chips. That left me with a lot more chips than before, but no one had been knocked out.
It was now 10:30 pm. Much to the surprise of the people at the table, I said I had to catch a plane so I was only going to play one more hand. ... So I look down at my cards and I have pocket aces! So I go all in like I would have with any two cards, and the guy next to me also goes all in with Ace-queen, thinking this is his lucky day. We turn over the cards and I beat him and I knock him out and now I have more chips than anyone. Which made leaving a lot harder because now I was the favorite to win the $3,000+ first prize. ... I announced that from here on out I was all in on every hand.
Following the same aggressive, all-in strategy he goes on to win the tournament (and catch his plane). He credits something called a commitment device, "which is when someone locks himself/herself into a course he/she wouldn’t otherwise want to have to follow, but as a result the person benefits." What he's saying is that usually he wouldn't have gone in on every hand, because that's a bad idea. But is it?
On the surface, yes. Knowing when to fold a heinous hand is a big advantage. It only takes one beat to lose it all. I'm certainly no expert, but I think there are two things going on that he doesn't take into account.
First, late in the game there are fewer players, and so hands don't have to be as good to win. While a mid-range pocket pair will almost certainly be busted with eight opponents, it stands a much better chance of holding up against three.
Second, a hand is much more valuable when you bet it than if you wait and call with it, especially late in the game, when the blinds are large compared to the stacks. Betting narrows the field, increasing your odds. Calling your bet is so expensive that many people aren't willing to take the risk, even if they have a better than average hand, because the only information they have about yours is that you bet with it. You're forcing your opponents to make a decision when faced with a perceived disadvantage. When the blinds are large, making people fold pre-flop can bring in a steady stream of income.
When you're big stack, both of these advantages increase. You should be aggressive, because even if someone goes all-in against you and wins, you're still in the game, and, if your stack is big enough, you're still in the lead. When your opponent's choice is between folding and going all in, they're going to wait to bet until they have a better hand than they would if you did not force them into that choice. Meanwhile, you get to rake in the blinds.
Should you go all-in every single hand? Of course not. Levitt got lucky, and I bet all the people he beat went home complaining about the idiot that took their money. Not that that's bad. Luck is part of the game, and you can't win a tournament without it.