Jae, I had just noticed the post after I posted this thread and then edited my post to include your link. I didn't have a chance to read it until now.
There's a few points of contention I have with the article. I may give some more thought to it and post to the article. The issue is related to risk/reward.
You can perform an experiment where you give people two alternatives:
Alternative 1: You get $100.
Alternative 2: We flip a coin. If it lands heads you get $200 and if it lands tales you get $0.
Most people tend to be risk adverse so they choose Alternative 1. The expected values are the same (assuming the coin is "fair"). Then we can modify Alternative 2 to change its expected value so that it is larger and see how much of a difference it has to get a person to choose Alternative 2 over Alternative 1.
Or consdier the following game:
If you don't play, you don't win or lose anything. If you do play there are two possible outcomes:
Outcome 1 – There is a 99% chance you will have to pay $100,000.
Outcome 2 – There is a 1% chance you will win $100,000,000.
The expected value is .99*-$100k + .01* $100m = $901,000.
I would conjecture that many people will not play this game if you could only play it once. At the very lest, I wouldn't. But if I could play it a large number of times, I would play it.
If you perform the event once, your odds of losing are 99%. If you perform the event 100 times, your odds of losing are only 36.6%.
There's perhaps a relationship here between risk and the benefits of diversification and the relationship between risk and long-term investing (why the standard finance view recommends high risk for younger individuals for their retirement and lower risk for those nearing retirement.)
Jae asked, "From a different angle, how can you measure something that is uncertain?"
Statistical analysis would be the short answer. The long answer is that it's going to depend upon what you're attempting to measure. Obviously you need sample information. From that you can calculate things like sample mean and sample standard deviation. Typically results are then assumed to be normally distributed and then you infer what the distribution looks like. It's certainly not a perfect way to go about things but it's better than going in blindly.
Casinos are based largely on this. They don't know what's going to happen on any individual occurrence (someone could win the jackpot). What they do know is what the distribution is for the games they have and they know the expected values. And provided that many games are played over and over again, the results will converge to their expectations.
Home
