杏吧原创

RUSSELL Crowe will win an Academy Award for his portrayal of mathematician John Nash in A Beautiful Mind. The film鈥檚 director, Ron Howard, will win the award for best director. Ian McKellen will take Best Supporting Actor. But Judi Dench can forget it.

Says who? Says a new way of predicting the future with a stunningly impressive track record. From film awards to presidential elections to scientific and technological advances, it has picked winners with uncanny reliability.

And where do these amazing insights come from? People like you. Those predictions of this year鈥檚 Oscars, for example, come from thousands of ordinary punters who are interested enough in films to play a Web-based game called the Hollywood Stock Exchange (HSX). Players who reckon they know who鈥檒l win the Best Actor Oscar can buy shares in him using fake 鈥淗ollywood dollars鈥�. The higher the demand, the more they鈥檒l have to fork out. But it could be worth it: come Oscar day, players with a big investment in the winner stand to gain vast amounts of鈥攚ell, nothing much, apart from kudos.

Playing for fake money on pretend stock markets might sound like a sad pastime for wannabe stock market traders (see 鈥淧laying the markets鈥�). But here鈥檚 a thing: those imaginary stock prices have turned out to be eerily accurate predictors of reality. Last year, HSX鈥檚 army of amateur traders correctly predicted the winners of six out of the eight main Oscar categories. The year before they nailed all eight鈥攂etter even than a poll of the Academy members who actually decided the winners. This year, too, the HSX traders seem to have lost none of their acumen: they correctly forecast 35 of the 40 Oscar nominees announced last month.

It鈥檚 a success rate that has led to the game being taken very seriously by movie moguls keen to gauge market reaction before they commit millions to a film. 鈥淲e get calls from agents and producers wanting their star, film title or whatever listed on the exchange,鈥� says HSX vice-president Michael O鈥橰orke. Some studios are even willing to pay HSX for detailed trading information.

HSX鈥檚 predictive power has also prompted people a long way from Tinseltown to wonder exactly what鈥檚 going on in these games. Are their forecasts really as good as they seem? An international team of scientists is on the case. What they are finding casts intriguing new light on some basic tenets of economic theory. And it could have a huge impact on how businesses carry out market research鈥攁nd how we consumers decide what we鈥檒l buy.

Not bad for something that started out as just a teaching aid. For years, economists have tried to find ways of convincing students that their theories bear some resemblance to reality. For example, the 鈥渆fficient market hypothesis鈥� says that share prices reflect the sum of all the knowledge available to every trader鈥攊ncluding the sneaky insider dealers. That鈥檚 because if they want to benefit from their information, they will do so through trading which changes the share price and thus reveals the implications of what they know to everyone else. Then a second basic tenet of economic theory kicks in: 鈥渞ational expectations鈥�, which says that everyone in the market responds rationally to what they know, and buys or sells accordingly. So stock markets allow communities of people to benefit from each other鈥檚 insights, with share prices giving a handy summary of what that knowledge implies.

All very neat, but is it true? In 1988, economists at the University of Iowa set up an artificial stock market to let students see the effect of some of these key ideas for themselves. The market traded securities, contracts that pay off only if an event takes place. This time they were tied to the candidates in that year鈥檚 presidential election.

Each security had to be paid for with real money. Once the election was over the securities paid off according to the percentage of the final vote won by each candidate. With real money at stake, students had a big incentive to follow the intricacies of the presidential race. And, according to the efficient market hypothesis, the final prices should have reflected the total knowledge of all the players. How did they do? The market predicted George Bush鈥檚 margin of victory over Michael Dukakis more accurately than any of the six national opinion polls.

Now, because of the Web, the Iowa market has opened its doors to anyone who cares to take a punt on political and financial events such as the New York mayoral elections or the financial performance of IBM. Renamed the Iowa Electronic Market, it has consistently beaten the major opinion polls in predicting the vote in presidential elections, including the dead heat in 2000.

The Iowa Market has spawned many imitators and trading is now possible on artificial markets covering a stunning range of issues. Formula One Pick Six, for example, trades in securities predicting the top six drivers in Grand Prix races, while NewsFutures trades in current affairs securities, such as the capture of Osama bin Laden by US forces. The Foresight Exchange even trades in predictions of scientific breakthroughs, from the likely date of a cure for AIDS to whether there will be a big quake on the US West Coast before 2010.

These artificial markets are played for imaginary money and usually offer little reward to successful traders apart from bragging rights. Yet they have proved hugely popular: HSX has over 400,000 accounts.

It would be all too easy to dismiss what happens in these markets as just a bit of fun, since with no real money at stake, no one鈥檚 going to fret that much over trading decisions. That should diminish their reliability. After all, people can make rash trades without worrying about losing their shirts.

But it seems that economic theorists forgot that there鈥檚 more to life than money. Research by a team led by Dave Pennock of the NEC Research Institute in Princeton, New Jersey, has revealed that people playing these toy markets do indeed act as if they have something to lose: self-esteem and peer-group approval. This triggers both market efficiency and rational expectations, which combine to produce knowledge pooling鈥攚ith a positive effect on the games鈥� predictive abilities. 鈥淧eople playing a market game are genuinely interested and care about doing well,鈥� says Pennock.

To find out exactly what was going on, Pennock and colleagues took a closer look at the movie market HSX. They found that its prices displayed a classic attribute of a rational, efficient market, called coherency. As the price of one Oscar nominee rose, those of at least one of the others dipped accordingly, keeping the sum of all the prices more or less constant.

Price coherency doesn鈥檛 just imply a market is operating efficiently and rationally. It also means prices can be used as estimates of something else that always adds up to the same figure: probability. In other words, the market won鈥檛 just tell you who will win, it will also tell you what their chances are.

Analysing over 100 prices quoted on the HSX, Pennock and his colleagues found that its price coherency was not quite as good as a real-money market. Its predictions remain pretty accurate, though. If Sissy Spacek鈥檚 shares are worth 40 per cent more than those of Judi Dench, then Spacek has a roughly 40 per cent bigger chance of taking the Best Actress Oscar.

Pennock鈥檚 team has uncovered price coherency in other artificial markets, including the Foresight Exchange, which deals in securities on all manner of future events ranging from Arnold Schwarzenegger becoming president to whether the Higgs particle will be found by 2005. Again the prices could be turned into probabilities. For example, a bunch of securities consistently trading at FX$0.80鈥�80 per cent of the maximum price鈥攃ame true around 80 per cent of the time. Similarly, claims with low prices, such as 鈥淥. J. Simpson vindicated by 2000鈥�, duly turned out to be true only very occasionally.

Not everyone鈥檚 impressed. Bob Worcester, chairman of London-based opinion survey company MORI and doyen of pollsters, dismisses artificial markets as just the latest in a long, sorry list of claims to have found a way of beating opinion polls: 鈥淲e must get half a dozen offers to go into business with these each week, and we always show them the door.鈥�

Worcester鈥檚 big problem with artificial markets is their lack of a mathematical basis. He points out that opinion polls are underpinned by tried-and-tested statistical theory, which demands them to be as large, random and unbiased as possible. The theory also allows estimates of the likely error: around plus or minus 3 per cent for the standard 1000-person poll. There鈥檚 no guarantee that an artificial market will be as rigorous, since the traders are usually self-selected and their numbers can be very small. Without this assurance, the results are beset by suspicions of bias and outright manipulation.

Yet opinion polls hardly have a glowing track record. The polls for the 1992 general election in Britain notoriously failed to get within 8 per cent of the actual result, chiefly because voters refused to give their views. But according to Worcester, alternatives such as artificial markets are even worse. 鈥淭hey鈥檙e voodoo polls,鈥� he says.

Pennock himself accepts that there are some key issues to be resolved鈥攏ot the least being how many active traders the markets need to produce accurate predictions. 鈥淚鈥檇 guess something on the order of 20 to 30 people could be enough,鈥� he says. 鈥淏ut the Web can draw from a global audience, so even esoteric topics could have enough to drive forecasts.鈥�

See the future

But one person at least is ready to bet the farm on the power of artificial markets. Later this year, Canada-based entrepreneur Leonard Brody plans to launch Ipreo, the first company offering artificial market services to those who want to see the future.

According to Brody, businesses all over the world spend billions of pounds a year trying to second-guess the response of real markets to uncertain future events. He thinks he can grab a big chunk of this from market researchers and focus groups by offering them an alternative with a proven track record. His artificial market would act as a financial test bed where real-life players trade according to how they see a company succeeding in the prevailing market conditions.

For example, private companies could track the performance of imaginary shares before dipping their toes in a real stock market. It could also help investors estimate a sensible valuation for a new company offering products or services never seen before on real markets. 鈥淭he Web allows us to tap into collective knowledge as never before,鈥� says Brody. 鈥淎rtificial markets capture this in a single metric鈥攑rice鈥攊nstead of the complex responses you get from market research.鈥�

Ipreo will initially draw its traders from business schools, offering them prizes as incentives for taking part. Brody recognises the dangers of market manipulation, and says that a raft of safeguards such as Internet address-checking will be put in place. 鈥淲e know who all our traders are, unlike stock markets like NASDAQ.鈥�

But it is not just companies that stand to benefit from this new source of insight, says Pennock. Artificial markets could also transform consumers鈥� decisions about what they buy. 鈥淚magine if you knew that VHS would win the video wars,鈥� he says. 鈥淵ou鈥檇 never have let the salesman talk you into buying that Betamax machine.鈥�

You can鈥檛 hear about trading on artificial markets without possible candidates springing to mind: the success of 3G mobile phones, for example, or the chances of nuclear fusion power before 2030. Oddly, no one seems to have set up the most obvious artificial market of all: shares in the success of artificial markets. No doubt it鈥檚 only a matter of time, and on the evidence to date, they鈥檙e a definite 鈥渂uy鈥�.