The Drunkard's Walk: How Randomness Rules Our Lives was a great read; I finished it in just about two days. It talks about the development of the sciences of probability and statistics and offers many examples of the role chance plays in our lives. I was quite taken with the thought that we all tend to overemphasize the role a CEO or a coach has in the success or failure of a company or team.
I have friend David Cramer to thank for the chance to read this book. He found it randomly in a coffee shop somewhere.
I drew the following conclusions from this book:
- The one thing you can control about your success in any complex endeavor (publishing, for example) is persistence. Success often goes to the person who won't give up, because no matter how good you are, you're likely to experience plenty of failure.
- If things go badly for you, rejoice in the fact that they're likely to get better as your circumstances regress towards the mean.
- Accept the ups and downs of life.
- Don't judge people too well for their successes or too harshly for their failures, as it's not necessarily indicative of their ability.
- Cheap wine is just as likely to taste good as expensive wine.
It had a number of fun examples and problems. My favorite started with this question:
If a family has two kids, and you know that one of them is a girl, what are the odds the other is a girl?
(The answer is 1 in 3)
Now if I tell you that one of them is a girl named Florida, does that change the odds, and how?
(The answer is yes, it does, and it makes the odds 1 in 2, and I had to read the explanation twice before I believed it.)
Highly recommended.
People dumb enough to name a child Florida are more likely to have daughters? Is all the information needed to answer this question stated in the premise, or does it rely on extra demographic data?
ReplyDeleteI did some more work on the Florida problem, I'm still not sure I believe it. It seems to hinge on the fact that knowing one more fact splits the probability set into more categories. I'm not sure whether I buy that or not. It seems intuitively wrong. But that's the logic. I did a spreadsheet but it relies on accepting some assumptions that I'm not sure about. The discussion in the book is interesting.
ReplyDeleteOk, here's something that should help with this. 1. The 'named Florida' thing is just supposed to be a detail about the known girl that limits the sample set. Florida is a rare name. 2. In the first problem, "how many 2-kid families with 1 girl have 2 girls", you only have to have 1 girl to be in the sample set; h. In the 2nd problem, "how many 2-kid families with at least 1 girl named Florida", you have to have a girl named Florida to get into the sample set, and if you have 2 girls to start with, you have 2 chances to have one named Florida, so 2-girl families are twice as likely to be in the sample set, and they end up being half the total.
ReplyDeleteThis last explanation makes some sense, I suppose. Having two girls would make it more likely that one of them would be named Florida. I guess the math says twice as likely.
ReplyDeleteThis is one of those hypotheticals that I would love to see tested against real world data.
Does the book say anything about people who just have crap luck with remarkable consistency? Or in given situations?
Part of the problem with statistics is that it bases conclusions on very large sample sizes. But in your own life, you're not going to keep trying certain things if you get really bad results the first few attempts. A few divorces early on can turn somebody off marriage; losing heavily at gambling your first few times can discourage you from gambling in the future. That seems like common sense. Also, some failures preclude future attempts: tear your shoulder up kayaking and you might not be ABLE to try that activity again.
I guess I'm saying that in real life, TIMING MATTERS, but not so much in statistics.
The book says nothing about folks who have lots of bad luck. However, in that circumstance, I think reversion to the mean is your friend.
ReplyDeleteRegarding large sample sizes and real life, I think about this a lot. Reading this sort of thing makes me believe that the rational way to live one's life is to utterly ignore your own experience in favor of the statistically averaged experiences of others, when making decisions. However, it's hard to believe anyone could successfully ignore his own experience, and one doesn't always have statistical data anyway. Still, it's an amusing idea: Yes, I got bit by a dog the last 4 times I went down that street. But most people never do...
It would be interesting to see how artificial intelligences approached this sort of scenario.
ReplyDeleteMost of the current research seems aimed at getting expert software systems or robots to learn from their experiences rather than basing their reactions on statistics.
I would think that in a free-flowing competitive situation, going with statistics would make an AI more predictable. I don't know that this would hurt it's overall winning average, but I would guess that when it lost, it would tend to lose big.