Until recently, there were approximately three people in the world who understood or cared about the theory of horse breeding known as Dosage. Even professionals in the thoroughbred business viewed the Dosage adherents as members of a lunatic fringe.

But as this Kentucky Derby approaches, Dosage has become a prime topic of conversation. The Daily Racing Form lists the Dosage ratings of all the candidates for the race. ABC-TV's Derby coverage Saturday will include a segment about the theory. Horsemen involved in the race now talk about it as matter of factly as Jeff Lukas, the assistant trainer and co-owner of Badger Land, did this morning, "I'm hoping," he said, "that our Dosage doesn't let us down on Saturday."

The theory of Dosage has attracted such attention from the cynical racing community because of its phenomenal record in the Kentucky Derby. Dosage figures purport to tell which horses do and don't have the pedigree to win at 1 1/4 miles, and the numbers have been correct every year since 1929.

Last year, the figures ruled out the strong favorite, Chief's Crown. Dr. Steven A. Roman, the chief guru of Dosage, told The Washington Post that he preferred Spend a Buck and Stephan's Odyssey. His selections ran one-two.

In view of this record, even skeptics cannot completely ignore the Dosage numbers that say Snow Chief, the overwhelming favorite, will not be able to win Saturday. Roman likes the chances of Badger Land, Ferdinand and Groovy.

As complicated as it may seem, the whole theory on Dosage rests on one simple premise: Certain stallions are so influential that they consistently pass on certain qualities to their descendants. J.J. Vullier, a French military officer who originated this way of looking at pedigrees, termed these stallions chefs de race.

Dr. Franco Varola, an Italian breeder, advanced the theory by placing the chefs in five categories according to the characteristics they transmitted -- ranging from extreme speed to extreme stamina. But it was Roman, a chemist with a strong interest in racing, who sought "to reduce pedigree to mathematical terms."

Roman assigned points to a horse for each chef who appeared in the last four generations of his pedigree -- 16 points if the chef was his sire; 8 points for a grandsire; 4 and 2 for earlier generations. Then he separated the points for chefs who tended to impart speed and those who tended to impart stamina to their offspring.

If a horse had 14 speed points and 7 stamina points in his pedigree, his Dosage Index -- the ratio of these two figures -- would be 2.0. If a horse has a low DI, he figures to be able to run a long distance. No horse with a DI above 4.0 has been able to win the Derby -- even though many of them try every year.

As it has gained more attention in recent years, Dosage has generated much debate in the thoroughbred industry, where countless millions of dollars rest on decisions about who to breed to whom.

The newsletter Racing Update denounced the entire concept in an article titled "Overdosed on Dosage," and its editor, Bill Oppenheim, said yesterday, "This whole thing is a crock. It's full of logical holes. Who defines what stallions are chefs? Who decides which category they go in? Believing that there are dominant sire lines is one thing. Believing that they are exactly of the same influence in individual cases is another. There is just not enough correspondence to reality to justify sinking money into decisions based on Dosage."

Roman responded that his method does require some subjective judgments and frequent tinkering but said, "I'm trying to find a relationship between pedigree and performance, and this is the most accurate method I know of to assess the racing characteristics a horse will have. But the preoccupation with specific races makes me nervous. If the streak in the Derby is broken, people like Oppenheim will say, 'See, it's not worth a damn.' "

Actually, the record of Dosage in the Derby is probably a misleading way to judge the theory's overall validity. The Derby is a unique race because it tests horses' ability to go a long distance at such an early stage of their careers.

Because none of them can have the optimal amount of conditioning to go 1 1/4 miles, their pedigrees matter more here than in any other race. Last year, for example, the Dosage Index said that Chief's Crown couldn't win at 1 1/4 miles in the Derby. The colt had no trouble handling 1 1/2 miles later in the year when he was more seasoned.

Among the experts, the debate may go on endlessly about the intellectual validity of Dosage and its practical uses in breeding decisions. But there is little room left for debate about its usefulness at Churchill Downs on the first Saturday in May.