In this way, with only 16 bits of memory (16 ones or zeroes), a machine by information theorist Claude Shannon beat all comers. To call this "outsmarting" the humans is perhaps a bit of a stretch, but it is what Poundstone means when he applies it to different areas.

There are a surprising number of areas where a similar "outguessing" strategy can be fruitful. Rock, paper, scissors is a random game, but because most people deviate from true randomness, it is possible to have a strategy. ("A player who loses is more likely to switch to a different throw the next time," he writes.)

A less obvious field of application is multiple-choice tests. But here also the same principle applies - as long as the tests have been assembled by people, rather than computers with randomising algorithms. Because, in such a test, "there is more true-false-true-false alternation than in a properly random sequence", the second of four possible answers is most often the correct one, and that "none of the above" and "all of the above" answers are disproportionately likely to be correct.

The most fascinating part of the book is a series of chapters in which Poundstone shows how sets of fraudulent numbers in made-up accounts or massaged expenses fall prey to the same kind of analysis. People overuse ascending pairs (21, 43, 76), and over-avoid doubled digits (55, 88); they don't know that there are surprisingly uneven distributions of first and second digits in sets of numbers from disciplines including finance and sport statistics. For such reasons, there were good reasons to be suspicious of Enron's revenue figures and Bernie Madoff's investment returns long before those scandals broke.

Poundstone begins to stretch with chapter titles such as "How to Outguess the Lottery", which is really just about how to choose more random sets of numbers than other players so that you won't have to share the jackpot in the unlikely event that you win.

He is funny about business forecasters, but goes oddly wrong on internet passwords, recommending you use a random string of only eight characters - "RPM8t4ka", say - and construct a mnemonic to remember it. In fact, a passphrase of several words in a nonsensical order - like "correct horse battery staple", in the celebrated example - is easier to remember and harder to crack.

Eventually Poundstone puts your money where his mouth is, with a set of betting systems. He describes the kinds of football games on which you might be able to beat bookmakers' odds, and advises on how to play the office Oscars pool.

The mother of all betting systems, of course, is a stock-market investment strategy, and Poundstone offers one of those, too: a long-term plan based on price-to-earnings ratios and sitting out periods of volatility. He doesn't offer the disclaimer from financial services advertising that past performance is no guide to future returns. His strategy looks excellent when tested against historical data. Can we assume history will continue in more or less the same way?

That kind of question leads to the deeper issue Poundstone doesn't explicitly address: what exactly is randomness, anyway? Is it just a way of describing our ignorance of what will happen next, or is randomness somehow built into the universe? This was an important question, after all, for Bohr and his fellow quantum physicists.

Perhaps there is a general "behavioural penalty" for being human in an uncaring universe, so we'd all do better if we were more like outguessing machines. Let's play rock, paper, scissors. Ready? One, two, three: paper.