Publish date:

Before you put that deuce down on a horse, better have a computer handy

A gambler named George Smith, back in the 1880s, is generally credited with being the first to keep detailed records of the past performances of race horses. Since he used the information to outguess the people who wanted to bet with him, Smith prospered and became famous as Pittsburgh Phil. Soon, however, New York newspapers were beating him at his own game. Their experts published daily forecasts of the races, in which they selected probable winners by analyzing records, just as Phil had done, but for the benefit of the general public, or at least of their readers. By 1894 Pittsburgh Phil was out of the picture, and there were no less than 60 prosperous bookmakers operating daily at Brighton Beach track, handling about $3,000 apiece on each race.

Now, in Horse Sense (McKay, $5.95), racing history seems to have come full cycle, for the author, Dr. Burton P. Fabricand, has discovered that the general public—at least in sum total—is more nearly right in its judgment of horseflesh than the experts. There is a kind of vast, miscellaneous, unorganized gambling democracy at the track—"the spectacle of each person wholeheartedly and effectively acting in his own select the right horses to represent him." The conflicting opinions of thousands "clash, interact, and modify one another in such a way that the final ensemble opinion is a closer approximation of the truth than that of any single individual." So the favorites win in a third of the races, and while it is true that a bettor who invests only on favorites will lose money, he will, says Dr. Fabricand, lose less than one who bets on other horses.

The problem, as he sees it, is to know when not to bet on the favorite. The system outlined in the 247 pages of Horse Sense consists of rules and equations, drawn from computer studies, to determine when the favorite is a good bet and when it is not. Dr. Fabricand, a 42-year-old physicist at Columbia University, contends that the public is most often confused when the past-performance record of the favorite is similar to that of other horses in the race. To meet the difficulty, the author sets up rules derived from computer analyses of 10,000 races. He explains the rules in very murky prose whose meaning (at a guess) is that it is better not to bet. Still, the ordinary horse-player has never received such a tribute to his intuitive insight, mathematical genius and racing knowledge as Dr. Fabricand and his computers have given him. It would surprise Pittsburgh Phil. It would even surprise most horseplayers.