Can a computer beat the races?
Increasing numbers of horseplayers are asking that question, wondering whether modern technology can solve the riddle that has baffled men for centuries. And increasing numbers of computer experts are taking an interest in the track, thinking that it may be a fertile area to apply their skills.
It is very doubtful that many of them will be able to use a computer to make a steady profit at the track. But this new technology will almost certainly alter and streamline the way handicappers approach the sport.
The pioneer of the computer era in handicapping was William Quirin, a mathematics professor at Adelphi University and an avid racing fan. Quirin fed data from thousands of races into a computer and then asked it all the fundamental questions. How important is a recent race? How important is a drop in claiming price? How important is early speed? Quirin found the mathematical probabilities for thousands of different handicapping situations, and published his results in "Winning at the Races: Computer Discoveries in Thoroughbred Handicapping." That book became the bible for the new generation of computer handicappers.
Many of them view computers not as a potential source of ultimate truth, but as a very valuable research tool. "They are an adjunct to the intellectual process," said Dick Mitchell, a Los Angeles mathematician. "They are to a handicapper what a word processor is to a writer." And many horseplayers are using computers in ingenious ways.
In Omaha, George Kaywood has programmed a computer to compile speed figures at his hometown track, Ak-Sar-Ben. In New York, Steve Scharff has programmed a pocket-sized calculator to analyze the pace of races with a method too sophisticated for pencil-and-paper calculation. But the most productive use of computers may be to study the performances and methods of trainers.
A young man named Eric Goldreich introduced himself to me at Santa Anita one day and asked if I would like to see some of the trainer research he had done. The next day, he came to the track lugging a book of computer printouts about three feet wide and about 300 pages thick. It looked, I thought, like a rather impressive analysis of all the West Coast trainers. Then I looked at the label on the cover: "Richard Mandella." This was a computer analysis of just one trainer. "I've got 45 more books like this at home," Goldreich said.
Goldreich knew Mandella's winning percentage with first-time starters in filly maiden special-weight races (his specialty); Mandella's record with first-time starters entered for a claiming price (dismal); everything about the trainer. Of course, the drawback of this exhaustive research was obvious. Goldreich didn't have time to do much of anything besides program data into his computer; he didn't even have time to handicap or get to the track often. Because computer-oriented handicappers have to generate all their own information, they commonly become slaves to their work.
But the day will probably come when this information is widely available. "How would you like to sit in front of your terminal," Quirin asked, "and type in the name of a horse and get your speed figures, trip notes, information on his physical appearance, data on his jockey and trainer. That should be possible--but it would take an organization. I think eventually we're going to see a network of some sort built up that will form a massive data base and let private subscribers tap into a big computer."
Yet, to some people, even such an ambitious use of computers seems relatively tame. If computers can be programmed to master chess, bridge and backgammon, to put men on the moon, why can't they create a system that will produce a consistent profit at the track?
There are many people using computers to find the Big Answer, and they tend to approach the quest in fundamentally the same way. They feed into the computer as many quantifiable bits of information about horses as they can, and then ask the computer to weigh those factors in the way that will produce the optimal results.
"The equation might be 10 percent for recency, 5 percent for the jockey, 10 percent for the trainer, 10 percent for early pace and then--the bomber, the killer--50 percent for speed," said Dick Mitchell. "The reason is that speed is the most quantifiable. You throw all this together and come out with a number that tells you each horse's percentage chance of winning. Then you measure that against his actual odds."
Mitchell says he has come up with a couple of special situations that have produced profits in an extensive test on paper, but discovering the Big Answer is another matter. A Los Angeles man, Steve Carroll, can attest to that.
Carroll and a partner bought an $8,000 computer and for the last 18 months have devoted full-time effort to their handicapping research. "We spent a lot of time writing software, a lot of time typing in data," Carroll said, "but we still haven't been able to find a model that works." Some computer people, he said, create winning systems by creating a system and then testing it on the same data from which it was generated, but Carroll said, "That's a no-no. We generate our data on 1981 and then test it on 1982, and every time we'd come up with a loss."
One of the difficulties Carroll found was that the importance of various factors he was using seemed to change from year to year. This might surprise and disconcert the mathematicians who attempt to approach the racetrack as if it were a chess board or a bridge table. But it would not suprise experienced handicappers, who appreciate the diverse and dynamic nature of the game.
Racing is very different at Santa Anita and Charles Town. It can also be very different from one day to another. Maryland horseplayers know this better than anybody. Sometimes, the condition of the track will have a small impact on results. But when a rail bias appears at Pimlico, it becomes the dominant handicapping factor. Computer people need to deal with a neat, well-ordered universe and they can't deal effectively with such change.
Although he has not abandoned his project, Carroll is beginning to understand this fact of racetrack life.
"I know people have a lot of wild dreams about what computers can do," he said, "but I'd warn them they'd better think twice. If I had it to do over again and decide how I'd spent that year and a half, I think I'd spend the time going to the track and handicapping."