Bests & Worsts Reviews from Amazon

This book is just another example of why non-statisticians shouldn't write books on statistical subjects. It's best-selling status confirms P.T.Barnum's famous phrase. What the author does is use year-end prices covering 45 years to "prove" all sorts of useless things about various simple strategies centering around value investing. Forty-five data points, whether it be daily data or yearly data, is not enough (by more than at least a factor of ten!) to reach any significant conclusion about anything such as the financial markets, where any signal present is swamped by noise. But O'Shaughnessy is quite tickled that he's proven all the experts wrong with his measly sample. What's even worse is that the spuriousness of his "profound" results is right there staring him in his face: he frequently comes up with numbers like 14.59%, with a standard deviation of 23.24% -- not realizing that this means the real answer is likely anywhere between -8.65% and +37.83%; in other words, it's consistent with a 0% return for that particular strategy. To compound the folly, O'Shaughnessy then goes on to continually draw conclusions based on differences of a few percent or less (a small fraction of the standard deviation) between competing strategies, not realizing that this is utterly meaningless. On top of these fatal flaws, O'Shaughnessy also commits the most common sin of model testing: he fails to forward-test his strategy on data which the model hasn't yet seen. This just about guarantees that future results will disappoint, as indeed has been the case the last several years with the mutual funds which he launched based on the ideas in this book after its first edition. I give it 1 1/2 - 2 stars because even a crummy strategy is better than none at all.