Crisscrossing telephone wires snaked along the carpeting of Rod Dufficy's cluttered home office near Hong Kong's Happy Valley racetrack. Dressed down in baggy black velour tracksuit bottoms and a matching gym shirt, Dufficy, 32 at the time, sat at a large L-shaped desk, rocking back in his chair and eyeing three computer screens crowded with numbers. He was cramming for a race that began in 22 minutes, calling up information from an online database and sifting it through a betting-analysis program built into his system.

The Australian was one of Hong Kong's elite breed of super-successful professional gamblers, computer-assisted horse bettors who work in teams and net millions at the races each year. That particular night, however, was not one of Dufficy's big-money sessions; he was down US$300,000 heading into this race. 'I'm going to have a lot of outlay on this next one,' he said. 'Maybe $550,000.'

Three slender Hong Kong sisters faced Dufficy, waiting for a printer to spit out a list of a few hundred bets with possible big payoffs. One sister grabbed the sheets and snipped them into strips, which she distributed to the other women. They punched the information into their handheld wagering machines, which transmitted it to the track via telephone lines, filling the room with the chirping of outgoing data.

Four minutes and three-quarters of a mile later, Dufficy looked up at a large-screen TV and watched thoroughbreds lunging past the finish line. He scrutinised a fistful of papers, then scanned the race results scrolling across his desktop monitor.

Strong in Hong Kong
'We've got wins - quinellas and a tierce,' he said calmly. He was $330,000 up on this single race. 'It'll put me ahead by $30,000 tonight,' he said. By week's end, his earnings would add up to a profit of about $130,000 for essentially two days of work - a typical cycle for Dufficy.

And there's plenty of dough to go around. So much so that Dufficy earned his fortune and recently left Hong Kong for his homeland of Australia. But it's not like Dufficy's departure has created a void or anything. Action still proliferates. As a half-dozen or so computer teams in Hong Kong will tell you, the allure centres on the city's massive handle - around $10 billion per year. It allows the teams to lay out hundreds of thousands of dollars on a single race without upsetting the odds.

But Hong Kong racing has other attractions as well: run by the not-for-profit Hong Kong Jockey Club, it is scrupulously honest (fixing would hurt computer bettors' calculations), and there is a pool of only 1,200 horses per season (a manageable number for the teams to track). Then there are the extravagantly exotic bets and parlays, comprising a rich smorgasbord of financial opportunities that seem custom-made for the computer teams. One, the Triple Trio, requires picking the top three finishers in three races and often pays six-figure dividends.

Computer teams pick winners by culling data from past performances and taking advantage of the public's miscalculations. They use bespoke software programs to determine their own odds, search for overlays, and place bets that can deliver big dividends for reduced risk.

Team leaders provide the multi-million dollar bankrolls, supplemented with investments from the 30 to 40 other members. Their jobs range from accounting to code writing to placing the bets. Annual salaries start at $50,000 for those who enter the wagers by phone and rise to more than $1 million for roles such as chief technology officers.

Westerners or Australians usually head up the teams and walk off with what's left after salaries and operating costs. William Ziemba, alumni professor of financial modelling at the University of British Columbia and a longtime observer of the scene, estimates that a top-notch outfit can pull in as much as $100 million in a good season, netting the boss a cool $50 million or more. While computer-generated horse-picking is not particularly new, it has reached an apex of sophistication in Hong Kong and is spreading beyond China. Teams have recently made inroads in the United States and Japan. You can easily spot tech gamblers at the track in Tokyo because they're the ones wheeling suitcases filled with yen.

Big fish, small bets
Competition among Hong Kong's computer teams is fierce. Tech secrets are closely guarded, nobody's keen to publicise their betting strategies and the cagiest players aim to hide their wagers from other teams - all of which monitor the flow of racing money via an independent online service called Telequote, based in Hong Kong.

Nobody's more skilled at masking bets than a team launched by Bill Benter, regarded by many as the most successful sports bettor in the world. Though he recently retired from racing, his team-members haven't. 'Normally, you'd see the odds go from 141/1 to 116/1 and know it's got to be a big professional bet,' says Dufficy. 'But Bill's team had its betting model set to disguise action with little $5,000 dribbles. They ultimately put the right amount on a horse, but did it over a sequence of time, leaving no footprints. That drove other punters crazy.'

Working from mathematical models that are calculated to deliver a 24% return on investments, Hong Kong's most sophisticated computer-assisted bettors operate with long-term certainty of what their profits should be. 'Racing is becoming more and more like a stock market model,' says Ziemba, who specialises in statistical analysis and edited The Efficiency of Racetrack Betting Markets, a collection of scholarly papers on the mathematics of horse wagering. Horses should be thought of as Microsoft or Dell, Ziemba says, and their past performances are the equivalent of economic charts that provide fodder for quantitative analysts.

'Racing is a financial market catching up with the rest of the world,' he says. The bedrock of a predictive betting system resides in a massive collection of data on each horse - including details about the tracks and jockeys. 'You massage all of that information into a mathematical equation that can be used for predicting probabilities,' he explains.

'If you wanted to get started in this, you would spend a year building the probabilities system, and it could cost $1 million to put together.' And that data bank needs constant updating.

Database driven
A typical team has employees whose sole job it is to review race tapes after every meet. They judge each horse on 130 characteristics - attributes like speed during the first third of the race, whether it got bumped coming out of a turn, the quality of its recovery from the bump, and how it finished - and assign numerical grades. This goes into the database, where it can be cross-referenced and called up to help predict the outcome of any impending race that particular horse runs in.

The software then determines each horse's likelihood of winning a race. When a horse's computer-generated odds are better than the public's odds, the team slams in its wagers.

'You create a model that can analyse each type of bet, judge the conditions (in terms of money in the pool and the associated odds), and tell you when it will be most favourable to bet,' explains Ziemba. 'You don't necessarily want to bet a ton every time - you only do it when you can find advantages.'

One top punter explains it like this: 'Our computer programme churns through the history of the horses and adjusts all the probability in a very sophisticated way. We feed that into our betting programme, which looks at all the odds for the various outcomes. It looks at your true chances of winning with the latest pay-off odds and calculates what the best potential bets are, based on the chances of winning and the odds. Then it runs through all the probabilities.

Formula Won Ton
'The mathematical aspect involves following a basic formulation that all successful gamblers use - whether they know it or not,' he says.

'It's having what mathematicians call a positive expectation on the bet. You multiply the probability of winning by the pay-off odds of one bet. Let's say the horse is 20/1. If it has a .05 probability of winning, you multiply that by 20/1. You get 1.0 - or 1/1 - and that is a fair pay-off bet.

'But if that same horse is paying 25/1, then it has a positive expectation. Now it is 1.25 (or 1.25/1). It gives you a 25% edge. Given that you know the true probability of winning, the amount to bet is a closed-form problem based on how much you can lay down without hurting your odds.'

Designing software to do all this is a delicate operation with seemingly endless pitfalls that can disastrously skew results. Benter, for example, went broke at least once before his system was efficient enough to turn a steady profit. 'Every year, more and more people come here and leave with their tail between their legs,' says Dufficy.

Whoever writes the team's software needs to decide early on which aspects of a horse's performance to take most seriously. For instance, if a debuting horse's odds of winning are 50/1 and it wins its first race, the software will note that - and might be inclined to view untried horses with long odds as good bets. So the system must be tweaked to give little weight to those outcomes. Other, more ambiguous factors - turf firmness, recent time trials, second-place finishes, and the jockeys' racing styles, to name a few - must also be taken into account.

'Memory is another thing,' suggests Kelly Busche, an economist who taught at Hong Kong University and consulted for one of the major teams.

'How quickly do you discount information? And to what degree? What happened two seasons ago should carry less weight than what happened last season. You need a model and a database that are both agile and robust enough to handle a variety of ever-changing situations.'

To build a good horse racing model, teams rely on workers with the skills of hedge fund technicians. Rumour has it that one of the teams has wooed programmers from Fortune 500 companies. 'You need a hardcore nerd who is good with numbers and has a mathematical and engineering background,' says one team leader. 'What we do with computers here is similar to what you see with Deep Blue. It's about attacking problems by fussing around and fine-tuning rather than using intuitive knowledge.'

Some punters believe the Jockey Club created big, complex bets like the Triple Trio as a hedge against the computer teams' skill-based advantages. The idea was that such wagers would be impossible to handicap, thus enhancing the luck factor and levelling the field. But things have not worked out that way. The biggest Triple Trio ever, paying a dividend of $18 million, was snagged by a pair of computer-assisted players who covered 900,000 possibilities with bets totalling $1.2 million. The top teams routinely make their fortunes through complicated parlays, quinellas, and exactas.

Revelation: the number of the beast
The story of computer-assisted betting in Hong Kong begins with Bill Benter, the US-educated, impeccably dressed technician who developed the first successful programme put to use at Happy Valley. The importance of his pioneering work is confirmed by rivals and experts alike. Benter got his start in the mid-1970s, when he discovered Beat the Dealer, a bible for blackjack card counters. He memorised the best-selling book's strategies and hit the casino circuit, where he met his future partner, Alan Woods, a former actuary turned counter. In Las Vegas Benter stumbled on a slim handicapping guide - and turned from casinos to horse racing.

Equipped with a $150,000 bankroll provided mostly by Woods, the two card-counters planned to apply the theories of winning at blackjack to winning at the races. Beat the Dealer, after all, had been written with the aid of a computer that analysed every possible situation at a blackjack table and assigned numerical values based on which cards remained in the deck. The idea, when you follow that best-selling guide, is to rigorously stick to its formula and bet high - even when you have only a tiny advantage. In the long run, despite frequent fluctuations and potentially long periods of losing, you will win a prescribed percentage of money.

By the time their computer programme had been fully refined, he and Woods had bitterly fallen out. But in the end, each wound up with an odds- and probability-crunching machine. Woods, now operating out of Manilla with a Hong Kong-based team, uses off-the-rack Pentium computers, still runs DOS, and employs an out-of-print program called Revelation for his database. At its core, it remains the original system.

Most telling of all, though, when winners cross the finish line, you don't hear even a whoop from Woods and his crew. As any punter can tell you, the real miracle of this technology is that winning fails to come as a surprise.

Exacting exactas? Exactly
The Hong Kong model is catching on. Consider that as the horses bolted from the starting gate at Gulfstream Park in Hallandale Beach, Florida, a few years ago, the odds for the winning horse suddenly went from 10/1 to 8/1. A gambler reported the shift to Gulfstream's Vice President of Finance, Bob Zambreny Jr, who discovered that in a three-second span, 167 exacta wagers were placed from a North Dakota betting outfit that attracts sky-high gamblers by taking low commissions and offering flexibility. Further snooping uncovered a computer team operating in the US using handicapping software similar to the systems employed in Hong Kong.

The team-leader used a special interface allowing him to batch his bets - to place dozens of wagers per second - into the system. This allowed the team to lay down a skein of complex exacta wagers after nearly all wagers were in, the odds were practically set, and it was relatively clear as to how much could be staked before upsetting the odds beyond a tolerable degree. The bets, automatically placed by the computer, ultimately paid out $246,020 on a total bet of $25,569. More impressively, over a 50-day period the team had reportedly netted $3.3 million in profits on $12.9 million in bets.

Gulfstream quickly barred the team from placing bets by computer, asserting all punters must have an even chance. Other racing professionals are more open-minded. In fact, Barry Schwartz, New York Racing Association chair, welcomes high tech action: 'Computers are simply another tool for handicapping. There's no guarantee of winning just because you're using one. These gamblers are willing to risk money just like everyone else.'

No guarantees, but the evidence from Hong Kong suggests it's the closest thing to a sure bet. There, the models have been taken as far as they can go, and it will be up to a new generation to tweak them for use in other locales. As handicapping systems seep into the mainstream, the innovators anticipate a future involving artificial intelligence - a Kubrickian computer that blends the human sensitivity of the best old-fashioned gamblers with the brute force of a supercomputer. If that day ever comes, traditional horse bettors may really have something to worry about.