Scan barcode
leogulus's review against another edition
5.0
bob_muller's review against another edition
5.0
My only criticism is purely personal--I would have liked to see more math. That would, of course, doom the book for popular readers. And yet, isn't that the problem? If you are in finance, and you have to "turn over the business to the quants" because you don't understand the math, then you're part of the problem.
Rant on.
I must also say that this problem is not limited to economics. I went to MIT in 1976 to study mathematical modeling with the most sophisticated professor out there doing it, and he published virtually nothing useful. The rest of the department there, while excellent at history and sociology, were limited to the more standard "statistical" models, which are usually simpler than those found in economics. I got my PhD and abandoned the field to go into software development, which was much more rewarding in many different ways. I recently had occasion to do some personal research that took me to the library shelf holding the books on mathematical modeling in Political Science. There was one book that I had not read by 1980, and that was by the guy I met on the first day I was at MIT, who got into agent-based modeling and made something of it.
There is an even more serious problem that Weatherall simply doesn't mention: data. I spent the last 11 years working with reference genome data--the structural and functional data around the "standard" genomic representations for various species (human, mouse, fruit fly, zebrafish, arabidopsis plants). I was a founder at a company which has the mission of developing a sustainable business model for such scientific data. What happens when the government stops funding a big database project? The data disappears because nobody will pay for its maintenance and development. Just before I joined this project, I worked for a year in a political science department. I found almost no large databases worth anything, and virtually no recognition by the professors doing data-oriented research that having large, ongoing data maintenance efforts was a worthwhile endeavor--mainly because no one would pay for it. The best data effort I've found in social science to date is the Piketty data on economic inequality, and it remains to be seen how much development that data set will see. Finance has reams of data, but it is very limited in scope and availability, and beyond the basic price data, the quality is deeply questionable (unemployement? inflation? Sure.). Political science and other social sciences have almost no real data of any sort, which pretty much eliminates any real chance of effective mathematical modeling in those "sciences."
Rant off.
daaan's review against another edition
3.0
charliemudd's review against another edition
4.0
wilte's review against another edition
5.0
From the French Bachelier who figured out a lot of statistics on random walks at the beginning of the 20th century, only to be forgotten for decades, via Osborne who argued that rates of return were normally distributed, not stock prices, to Mandelbrot (of fractal-fame) who stressed fat-tails and non-normal distributions (but Cauchy / Lévy stable distributions).
Also includes Ed Thorp (of [book:Beat the Dealer: A Winning Strategy for the Game of Twenty-One|891883] fame), who started one of the first hedge funds; Fischer Black (Black-Scholes formula to price options); Packard and Doyne Farmer (The prediction company [later bought by the enigmatic and secretive O'Connor and Associates, using chaos theory, black box modeling and genetic algorithms, both worked at Santa Fe institute); Sornette who used prediction models for earthquakes to predict large financial crashes (and made 400% putting his money where his mouth was in the 1997 crash).
"The stories in this book show the [scientific/physics] methodology in action: one uses simplifying assumptions to make a problem tractable and solve it. Then, once you see how your solution works, you can double back and begin asking what happens when you play with your assumptions" (p209)
financial modeling is an evolving process, one that proceeds in iterative fashion. (...) Models fail. Sometimes we can anticipate when they will fail (...) in other cases, we figure out what went wrong only as we are trying to put the pieces back together. (...) since mathematical modeling in finance is an evolving process, we should fully expect that new methods can be developed that will begin to solve the problems that have plagued the models that have gotten us to where we are today. (p128-129)
Data outclass theory (p.155)
If you continue to trade based on a model whose assumptions have ceased to be met by the market, and you lose money, it's hardly a failure of the model. It's like attaching a car engine to a plane and being disappointed when it doesn't fly (p47)
no matter how good the theoretical backing for your (non-black box) model, you ultimately need to evaluate it on the basis of how well it performs. Even the most transparent models need to be constantly tested by just the same kind of statistical methods that are used to evaluate black box models. (p155-156)
bibliophile026's review against another edition
2.0
davidsteinsaltz's review against another edition
3.0
dav's review against another edition
4.0
It's a bit light, but a nice appetizer if you are interested in the quant phenomenon.