HALLANDALE, Fla. - Racing fans across the country have often observed dramatic, last-second fluctuations in horses' odds, giving rise to various conspiracy theories. Many suspect that bets are being placed after a race has begun. But few could have guessed the true nature of some large wagers that cause payoffs to plummet: They are being placed by a computer.
This revelation came to light after extraordinary betting action on the fourth race at Gulfstream Park on Feb. 16. The 1-11 exacta combination of Marketchase and Proud of It was worth about $1,100 on the next-to-last flash of the tote board, but after the horses crossed the finish line first and second, the exacta paid $703.60. At the same time, Marketplace's win price dropped from 10-1 to 8-1.
An out-of-town gambler observed this action and reported it to Bob Zambreny Jr., Gulfstream's vice president for finance. Zambreny launched an inquiry and discovered that large, winning wagers on the race had come from a North Dakota off-track betting operation, Racing Services Inc. This was no surprise. Some telephone betting services are suspected of offering rebates or profit-sharing deals to high rollers, and many of the nation's biggest and most skilled bettors wager through these outlets. Reportedly, a syndicate of gamblers was behind this big play.
When Zambreny obtained a detailing of all the wagers from Racing Services on the Gulfstream race, the data was a surprise -- and a revelation. The printout showed each wager and the precise time (to the second) that it was made. A small part of it looked like this:
2:25:38 p.m. $310 exacta, 10-4.
2:25:38 p.m. $217 exacta, 9-11.
2:25.38 p.m. $610 exacta, 3-10.
2:25:38 p.m. $321 exacta, 5-11.
2:25:38 p.m. $190 exacta, 1-11.
The latter bet was the one that caused the winning payoff to plummet. Half a minute later, more money was wagered on the same combination, and these exactas returned more than $71,000. The bettor also collected in every other pool -- win, place, show and trifecta. Overall, North Dakota bet $25,569 on the Gulfstream race and received $246,020 -- a profit of $220,451.
There was something even more extraordinary about these wagers than the size of the profit. Between 2:25:38 and 2:25:41 p.m. -- a span of three seconds -- a total of 167 separate wagers were placed from North Dakota on the Gulfstream race. Half a minute later, in a space of two seconds, another 95 bets were made. And the amounts were odd ones: Why would anyone place a $217 bet?
When Gulfstream officials looked at the printout, they deduced immediately that the bets must have been made electronically. Nobody with a telephone account can rattle off 167 separate plays in three seconds. When I looked at the printout, I knew what the computer (and its programmer) were doing.
In 1994 I visited Hong Kong, where I was permitted to observe the operations of the world's most successful horse bettor, who had programmed a computer to make his wagers for him. In a sense, Mr. B. did what all horseplayers strive to do: He estimated the winning probability of every horse, and bet when the odds were favorable. He used a computer to calculate the probabilities, but the key element to his method was determining how much to bet.
If a horse rated as a 3-to-1 shot, but went off at 2 to 1, Mr. B. would shun him. If the horse was 4 to 1, he would bet. At 10 to 1 he'd bet a lot. Ordinary handicappers would do the same. But how can you make such judgments about every exacta and trifecta combination in a 12-horse field? Only with a computer.
Mr. B. received a direct feed of betting data from the track; his computer scanned every exacta and trifecta combination, identified the ones offering value and then determined the proper size of a bet that would maximize profits without cutting the odds too sharply. "Betting strategy is a solved problem," Mr. B. declared.
The solution - the formula for optimal betting - is contained in a book, "The Theory of Gambling and Statistical Logic," by Richard A. Epstein.
Mr. B.'s computer specified the precise amount to be bet on dozens of combinations in every race - all the ones that promised to offer value. As a result, his wagers would regularly be in odd amounts -- such as $217 on an exacta. Once I might have thought such an approach was far-fetched, but Mr. B. had made a fortune in Hong Kong. He told me, "Somebody ought to try to do this in America."
Now somebody has. To employ this system in Hong Kong or America requires betting data (such as the dollar amount wagered on each exacta combination) so that the computer can perform its necessary calculations in the final seconds before post time. But almost no outsider has access to the data from Amtote, the company that services Gulfstream and many other U.S. tracks. Even Gulfstream itself doesn't have such direct access. But the group betting through Racing Services somehow obtained it.
In so doing, they secured a significant advantage over their parimutuel competition. Every horseplayer runs the risk of having his odds change after he has placed a bet. But the North Dakota bettors have received an exemption from this disadvantage. An outraged professional gambler observed that this is equivalent to "a secret auction where one bidder gets to see all the other bids and then decides whether he wants to participate."
When Gulfstream officials learned what the North Dakota group was doing, they concluded that its computer betting was unfair to their other customers, and ordered Racing Services to stop it. As of March 2, it was stopped - temporarily at least. Susan Bala, president of Racing Services, saw nothing wrong with the practice and told the Daily Racing Form that the computer wagers would be halted "until this can be worked out."
The rest of America's horseplayers should be outraged that one group of players was allowed to take such an edge. But they will also feel a grudging admiration for anyone shrewd enough to obtain that edge and smart enough to exploit it.
© 2001 The Washington Post Company