Fooled By Randomness
Occasionally, friends or co-workers ask me to recommend investing books. I used to suggest a list of reading, but now I'm going to change this standard advice.
If you can only read one book about investing, read Nassim Taleb's Fooled by Randomness. If you can read more than one book, read it first, anyway, to put the other stuff in context.
This book has much wider applications than finance. He makes convincing arguments about the limits of predictions that are based on (a narrow interpretation of) historical data.
For instance the common complaint: "this has never happened before! It's a rare event! We could not have foreseen it!" Well, history is full of things that had never happened before happening. So the lesson is that time-series can't really be used to predict the future. You somehow have to consider how well you would do across all possible futures. If there's a high variance, then maybe your success is not due to anything you actually did, but rather to random events.
He also makes a convincing case that many practitioners of risk-management are working on fundamentally wrong assumptions.
I see some of the same patterns in software engineering and management. If a project succeeds in spite of gross mis-management, its leaders are praised as visionaries and risk-takers. Never mind that if they tried the same process 10 times, they'd fail most of the time. If a more thoughtful group follows a (meta) process that would work most of the time, but fails in a particular instance (due to random events), then they're easy targets for finger-pointing. Why were they so conservative?
I.e. if someone's playing Russian roulette for no good reason (or worse, doesn't realize that they are), we rightly call them foolish. However, in practice, for more complicated scenarios, we typically reward success that is often the result of random factors, not the actions of the agents we praise.
Now, Taleb is fairly arrogant, and makes sweeping generalizations, at times. If you're an MBA, just put on your flame-retardant underwear -- there's useful information here.
My wife pointed out that some of his statements about medical research are plainly wrong. For instance, that statisticians don't work closely with doctors. This makes me wonder about his other generalizations. The way I read it, he's a ranter who has been surrounded by arrogant people who talk the talk, but understand little. He's a product of that environment.
If you can only read one book about investing, read Nassim Taleb's Fooled by Randomness. If you can read more than one book, read it first, anyway, to put the other stuff in context.
This book has much wider applications than finance. He makes convincing arguments about the limits of predictions that are based on (a narrow interpretation of) historical data.
For instance the common complaint: "this has never happened before! It's a rare event! We could not have foreseen it!" Well, history is full of things that had never happened before happening. So the lesson is that time-series can't really be used to predict the future. You somehow have to consider how well you would do across all possible futures. If there's a high variance, then maybe your success is not due to anything you actually did, but rather to random events.
He also makes a convincing case that many practitioners of risk-management are working on fundamentally wrong assumptions.
I see some of the same patterns in software engineering and management. If a project succeeds in spite of gross mis-management, its leaders are praised as visionaries and risk-takers. Never mind that if they tried the same process 10 times, they'd fail most of the time. If a more thoughtful group follows a (meta) process that would work most of the time, but fails in a particular instance (due to random events), then they're easy targets for finger-pointing. Why were they so conservative?
I.e. if someone's playing Russian roulette for no good reason (or worse, doesn't realize that they are), we rightly call them foolish. However, in practice, for more complicated scenarios, we typically reward success that is often the result of random factors, not the actions of the agents we praise.
Now, Taleb is fairly arrogant, and makes sweeping generalizations, at times. If you're an MBA, just put on your flame-retardant underwear -- there's useful information here.
My wife pointed out that some of his statements about medical research are plainly wrong. For instance, that statisticians don't work closely with doctors. This makes me wonder about his other generalizations. The way I read it, he's a ranter who has been surrounded by arrogant people who talk the talk, but understand little. He's a product of that environment.
Labels: finance, project management, randomness, reason
posted by Jaba Adams at 11:32 AM
0 comments
