4 ms·
St. Petersburg paradox
- nostrademons 17y agoI would pay a large sum of money to enter this - specifically, up to $0.50/expected iteration, minus electricity costs - if I were allowed to write a computer program to play for me. That suggests that the real resolution to this paradox has something to do with transaction costs. I wouldn't pay more than a few bucks to play manually because I don't expect to have time to play more than a few times, so the low-probability high-payoff events will probably never happen. I would pay a lot more if I could automate it, because then my cost per round becomes negligible, and I can rely on pure math. This may be why most normal people don't found startups (they don't expect to have enough chances for their lottery number to come up), and why quantitative hedge funds can trade profitably (they can take advantage of arbitrage opportunities that most investors overlook because transaction costs make it not worth their while).
- roundsquare 17y agoThis is hardly a paradox anymore. It only appears to be one if you accept "maximize accepted utility" as the decision metric. Maximizing expected utility is a pretty good decision metric, but it has problems. Specifically, it has huge problems in handling very-low-probability-very-high-payoff outcomes. Here's a simple example: Game 1: You get $1.00 Game 2: 99.999% chance of getting nothing. 0.001% chance of getting $100,000.00 The expected payoff in each is the same but they are clearly very different. Note: I realize I'm equating money with utility here. Just doing it for simplicity, you can replace with whatever you want to get a utility of 1 and 100,000 respectively. Note 2: You can argue that utility is bounded (say between 0 and 1). But I don't think this solves the problem.