Belmont Park appears in the book as a site of evidence about collective prediction: its 1970 horse races, considered alongside those at Aqueduct, supplied a large body of outcomes against which betting odds could be tested. The results showed that a market’s combined judgments could closely anticipate both the order of finish and the probability that a horse would win.

Setting and Use in the Study

Belmont Park was one of two racetracks included in a study by Arthur Hoerl and Herbert Fallin of every race run at Belmont Park and Aqueduct in 1970. The study covered 1,825 races. Rather than treating betting odds as mere tips, the researchers compared the market’s assigned chances with what actually happened on the track, asking whether the odds predicted both win frequencies and horses’ relative finishing order.

Market Rankings and Race Results

With only a few exceptions, the market’s ranking of the horses predicted the order in which they finished. This pattern held across fields ranging from five to twelve horses: the favorite finished first most often, the second-favored horse finished second most often, and the same correspondence continued down the rankings. The average finishing position also declined monotonically with each lower market rank, with no mistake in that measure.

Odds as Probabilities

The study also tested whether the odds captured the actual frequency of victory. Among the 312 races with seven-horse fields, bettors’ odds predicted a 33 percent win rate for the favorite, which won 34 percent of the time. For the second-ranked horse, the predicted and observed rates were 22 and 21 percent; for the third, both were 16 percent; and for the fourth, both were 12 percent. The lower-ranked horses showed similarly close figures: fifth, 9 percent predicted and 8 percent observed; sixth, 6 and 8 percent; seventh, 3 and 2 percent. Results in races with other field sizes were similar, though not quite as striking.

Significance

Belmont Park’s races helped demonstrate how a betting market can turn many individual judgments into a forecast that tracks observable outcomes. The evidence is specific to the 1970 races studied, not a claim that every market or prediction is invariably accurate. Within that defined test, however, the odds at Belmont Park and Aqueduct came remarkably close to matching both the horses’ finishing ranks and their objective probabilities of winning.