Recently read: Fooled by Randomness
Fooled by Randomness, by Nassim Nicholas Taleb
This is a problematic book: He has a very important point, that much of life is a result of randomness more than skill, and to judge simply based on results can lead to catastrophic outcomes. The archetypal example is a financial trade that has a large probability of a small gain, and a small probability of a huge loss. Merely because a trader has been lucky so far doesn't imply continued future success: Eventually the loss will tend to occur, more than wiping out all the accumulated gains. (The pattern is common enough to have a name on Wall Street: "Blowing up") Furthermore, because the unlikely events are rare, it's exceedingly difficult to know how bad they'll be when they do occur. There simply isn't enough data to estimate with any confidence.
But the book is flawed, in that he sometimes oversimplifies, and sometimes exaggerates, and is sometimes mistaken. The cumulative effect is such that I really don't know what to take seriously, when so many of the details that I can verify don't add up. Two sloppy points come to mind: On page 206 he presents a problem of a medical test with a 5% false positive rate, and 1 in 1000 true incidence, and asks for the probability that, given a positive test, the patient actually has the disease. His point that it's unlikely is valid, but to be complete, we also need the false-negative rate, which he omits. In another example he conflates probability with confidence, which is admittedly a subtle issue, but one that a critique of improper use of probability should get right.
He's apparently very fond of Monte Carlo simulations, distrustful of pure mathematics, and probably doesn't understand chaos (but I can't be sure, he might be oversimplifying or exaggerating).
In short, I wish someone more careful had written a similar sort of book, so I could recommend that instead.
Not recommended.
This is a problematic book: He has a very important point, that much of life is a result of randomness more than skill, and to judge simply based on results can lead to catastrophic outcomes. The archetypal example is a financial trade that has a large probability of a small gain, and a small probability of a huge loss. Merely because a trader has been lucky so far doesn't imply continued future success: Eventually the loss will tend to occur, more than wiping out all the accumulated gains. (The pattern is common enough to have a name on Wall Street: "Blowing up") Furthermore, because the unlikely events are rare, it's exceedingly difficult to know how bad they'll be when they do occur. There simply isn't enough data to estimate with any confidence.
But the book is flawed, in that he sometimes oversimplifies, and sometimes exaggerates, and is sometimes mistaken. The cumulative effect is such that I really don't know what to take seriously, when so many of the details that I can verify don't add up. Two sloppy points come to mind: On page 206 he presents a problem of a medical test with a 5% false positive rate, and 1 in 1000 true incidence, and asks for the probability that, given a positive test, the patient actually has the disease. His point that it's unlikely is valid, but to be complete, we also need the false-negative rate, which he omits. In another example he conflates probability with confidence, which is admittedly a subtle issue, but one that a critique of improper use of probability should get right.
He's apparently very fond of Monte Carlo simulations, distrustful of pure mathematics, and probably doesn't understand chaos (but I can't be sure, he might be oversimplifying or exaggerating).
In short, I wish someone more careful had written a similar sort of book, so I could recommend that instead.
Not recommended.