## Math, Poker, and Perfectly Reasonable Deviations

The Math and Poker blog is one I like to go to periodically (not the least because I am part of his blogroll). The thing I like best is the use of simple probability/queueing/stochastic systems to point out the misconceptions of many gamblers. A little probability goes a long way, but it seems that many gamblers are not willing to even take the first steps in probability theory.

A recent post on that blog pointed me to Perfectly Reasonable Deviations, a blog on “random ramblings on science, technology, and other such stuff”. A recent entry there reviewed the book Convex Optimization by Boyd and Vandenberghe. An excerpt from his review:

It’s a wonderful book! A masterpiece! A joy to read! Quite likely one of the best math books I have ever read.

I agree!

## { 3 } Comments

Thanks for linking to my blog.

I am yet to find someone who does not LOVE that book! 🙂

Cool blog! I found it some months ago via “Math and Poker”. Interesting stuff on OR you have here. Unfortunately, OR seems to draw very little attention from both the scientific and the business communitites alike. It’s like an “enabling” field: it enables other people’s work, but remains relatively unknown to most.

Most gamblers think probablity stops with answering the question “What’s the probability that the next card will be a spade?” The ones that go past that understand expected value and know what variance is, but they don’t really understand variance.

I havn’t even finished the first chapter of Convex Optimization yet, but so far I really like it. The emphasis on the model rather than the solution algorithm is something that I don’t think is done nearly enough.

I agree that this is a wonderful book. Readers who are more interested in optimization methods than modeling may be disappointed though- the book doesn’t put a lot of effort into discussing how the problems are solved.

In fact, the book concerns itself primarily with convex optimization problems that have particular structure (e.g. second order cone programming, semidefinite programming, etc.) that can and should be exploited in the solution process. Solving more general convex optimization problems is more difficult, but the authors don’t really dwell on this.

A major challenge in this area is developing modeling languages to support the use of these techniques. Right now, it’s basically up to the user to do the transformations to bring his problem into a standard form. Ideally, a lot of this work could be done by a modeling language. The support for this in languages like AMPL and GAMS is very limited at this point.

## Share Your Views!