Commitments of Traders by Floyd Upperman

This book sucks. From the title and description, it should have an in-depth analysis of the Commitments of Traders report from the CFTC. As a self-professed engineer with a love of statistics (page 20), Upperman should have a great vantagepoint. Unfortunately, his engineering and analyses have major flaws.

Upperman either ignores or doesn't understand that successful engineering depends completely on physics. Without well-tested explanative hypotheses that describe how things work, engineers find it very difficult to put things together to fulfill design specs. Upperman skips building a corpus of well-tested hypotheses, and jumps into data mining. Having found some patterns that look profitable (methodology unstated), he back-tests them over arbitrary time ranges, and (surprise!) finds that they look profitable.

"Being right or wrong isn't the goal, the goal is to make money." (page 124)
Without a working model of financial engineering, Upperman finds himself making money on trades that he can't explain, i.e. trading by gut. Further evidence of his muddled thinking lies in his broken conception of risk management, summarized by the dictum "never allocate more than 10% of capital to any one trade" (page 192). This ignores covariance between positions, i.e. being long equities in Japan, Hong Kong, Taiwan, Thailand, and Korea puts 50% of your equity at risk of being zeroed if something affects Asia (like the recent decline in equities).

Floyd Upperman and his editor, Kevin Commins, at John Wiley & Sons either do not know enough, or they knew better and published this text anyway simply because they guessed that enough people would be conned. Either way, they should be roundly censured.

Mmm... black box trading. Sounds like fun... - David W