Steven Roman - Dosage Index, Leaving the game

[delayjf]

I was a huge Strike the Gold fan and he was my Derby horse.

The same reason I got off of Charismatic.

[Cratos]

Originally Posted by dilanesp

If you go to a roulette wheel, black or odd will at some point be "deadly" for a period of time....

It is a clear post hoc fallacy to assume that the reason Dosage no longer "works" is because of changes in breeding. The most obvious explanation is that Dosage was working with a very small sample so that there was no reason to think it was making statistically valid conclusions in the first place.

Your assumption that the change in breeding by the breeders’ didn’t affect the Roman Dosage methodology is a fallacious postulation and the Law of Large Numbers proves that assertion.

Yes, Roman’s sample size was small, but it was within the parametrics of the breeding scenarios at the time. Statistical extrapolation to large sample sizes doesn’t always give a better curve fit.

[Tom]

I jade a ton of money back int he day with Dosage.
There were only a few horses who met the 4.0 guideline and you could pretty much safely throw out the rest. That generally included everything Lukas showed up with early on. Nowadays, most everything qualifies...why is that?
Maybe breeders caught on?

Strike the Gold - yes, I threw him out.
But so what, you lose one bet. And that his numbers changed, so what - more data came in and it was updated. Beyer updates figures all the time.

Two of the most potent thing I ever used in racing were dosage and Tomlinson grass numbers.

[RXB]

Even in its earliest days he pointed out that the DI of stakes horses was rising and that stamina breeding was giving way to speed breeding. 35 years later, people are still taking that 4.00 thing to heart? That's their own fault, not Steve Roman's.

[dilanesp]

Originally Posted by classhandicapper

If you read the entire document, he addressed handicappers using Dosage as an eliminating factor for the Derby. It sounds like it annoyed him because that was not the intended purpose and he felt the misuse by others diminished the value of the work.

Imagine I created a class rating to help estimate the quality of a field relative to PAR and people started using it to arbitrarily bet on horses coming out of strong fields and against horses coming out of weak fields. It would fail. The fact that it failed what it was not designed to do does not mean it wasn't doing what it was intended to do really well.

Class, I think he's partly responsible for what he calls the "misuse" of his work for a couple of reasons.

First of all, he never poured cold water on it. I was around when everyone was talking Dosage in the 1980's, and Roman was a minor celebrity in horse racing. He enjoyed it. As long as the system was winning, he wasn't out there saying "hey, don't use this for handicapping".

Second, it isn't really the truth that the only problem was his system was "misused". In fact, it wasn't a statistically valid study to begin with. There simply are not enough Kentucky Derbies to make ANY statistically valid conclusions about sire influence. And the Derbies that have been run are not remotely a uniform, controlled sample-- for example, the 1973 Derby, with a short field and a superstar winner, is completely different from the 1974 Derby, with a gigantic field and a mediocre longshot winner.

So he was purporting to make "scientific" and "statistical" conclusions about things that he had insufficient warrant to make conclusions about. Maybe he didn't explicitly say "go use this to bet", but he did say "this is meaningful" and it wasn't.

[dilanesp]

Originally Posted by Cratos

Your assumption that the change in breeding by the breeders’ didn’t affect the Roman Dosage methodology is a fallacious postulation and the Law of Large Numbers proves that assertion.

I didn't assume anything. When you are dealing with too small a sample size, the default conclusion is that any changes in the trend line are due to variance, and the people claiming that it is something else have the burden of proof.

[Grits]

Originally Posted by PaceAdvantage

I like how he took his parting shot at Beyer...called him Andrew instead of Andy as well, like he's his dad....jealousy runs deep and long....

In all of its negativity, this too, is the feeling I got. But, I felt this about the entire piece. His boredom, his bitterness, outweighed any plus, even any gratitude that could have been realized through (and by) his contribution to the sport. Maybe he'll feel better after reading some of the posts that you gentlemen have written. He won't, at mine. But its OK. Obviously, this man had a good run. I hope he'll remember this.. JMO

[Tom]

Burden of proof?
I think not.

We have the risk of our bets using it and the pleasure of the profits we realized in doing so.

[dilanesp]

Here are some other examples of things that are wrong with Dosage:

1. He assumed there was something special about the Kentucky Derby. But that's silly. Snow Chief did have terrible Dosage, and did run up the track on Derby Day. But he came back 2 weeks later and beat most of the same horses at 1 3/16 miles, just 110 yards shorter, at Pimlico, and then 2 weeks after that won the Jersey Derby at 1 1/4 miles. The next year, he beat the 2 other best horses of his generation, Ferdinand and Broad Brush, and the best other Cal Bred of his generation, Variety Road, going 1 1/4 miles in the Strub. In other words, on no planet did Snow Chief lack the stamina to go 1 1/4 miles. He was beaten in the Derby because he was too close to a fast pace. Had nothing to do with his Dosage. But I remember Roman crowing at the time about it (Snow Chief was a Derby favorite and he said he was a toss-out).

2. He attributed way too much influence to breeding with respect to stamina. It is of course true that over a large number of horses and many generations, some bloodlines transmit stamina ON AVERAGE. But that is only one factor, and not really a controlling one, with respect to specific horses. History is littered with horses who shouldn't have been able to get a distance and did. Snow Chief, by the California sprint sire Reflected Glory, who I already mentioned, was one of them. Here are a couple of others:

a. Native Diver, one of the best horses in the history of California racing, had sprint breeding. He was by Imbros from Fleet Diver, two speedy California stakes winners. Native Diver won the Hollywood Gold Cup at 1 1/4 miles three years in a row, one of only two horses to ever do that. He was not supposed to get 10 furlongs.

b. Creme Fraiche had sprint breeding-- and, by the way, lousy Dosage. He was by Rich Cream, a sprinter who once set a world record in a sprint race at Hollywood Park. Creme Fraiche won the Belmont Stakes and later the Jockey Club Gold Cup when it was 1 1/2 miles. In the latter race, he beat Java Gold, who was by far the best horse in New York at the time. During the same period Roman was crowing about Derby favorites, Creme Fraiche was a continuing embarrassment to him.

The reality is that there really just isn't a lot that breeding can say about whether a horse will get a particular distance in a particular race. It depends a lot on trip, pace, training, and a lot of other things other than breeding. And Roman, obsessed with the Derby (as so many racing fans are), got way outside the realm of statistical validity and got a lot of publicity for himself peddling a bogus theory.

[whodoyoulike]

Originally Posted by Tom

I jade a ton of money back int he day with Dosage.
There were only a few horses who met the 4.0 guideline and you could pretty much safely throw out the rest. ...

I know very little about D.I. So, what were the D.I.'s of the other 1,000's of other horses which raced in the TC series. That, I would think would have increased the size of the sample population.

Were the winners the only ones with 4.0 or less?

I didn't understand that part.

And, why is it only effective up thru the TC series instead of all the other races being run e.g., Beyer, Sheets, Thorograph etc.?

I remember he was posting on here last year maybe he or someone familiar with the process can provide me with a better (but a little late) insight. And, this changing the numbers after other facts come up I just don't know about that part.

[Cratos]

Originally Posted by dilanesp

I didn't assume anything. When you are dealing with too small a sample size, the default conclusion is that any changes in the trend line are due to variance, and the people claiming that it is something else have the burden of proof.

Dr. Roman’s Dosage methodology appear to have worked on frequency statistics which works on the number of occurrences of an attributes and there is not case of dependency, but are you suggesting that he should have used. Bayesian statistics which are applicable in the data sets where every attribute is dependent on every other attributes and every attributes existence is purely conditional?

[dilanesp]

Originally Posted by Cratos

Dr. Roman’s Dosage methodology appear to have worked on frequency statistics which works on the number of occurrences of an attributes and there is not case of dependency, but are you suggesting that he should have used. Bayesian statistics which are applicable in the data sets where every attribute is dependent on every other attributes and every attributes existence is purely conditional?

I'm really suggesting that given both the small sample size and the lack of uniformity in the sample, there's really nothing that can be said about the influence of sire lines in the Derby specifically that's going to be very useful.

What COULD be useful is a Dosage-like system that looks at ALL 1 1/4 mile races, or all races of longer than a particular distance, if it could be statistically established that there was a significant correlation between influential sires in the pedigree and performance in such races. But such a system would only have limited utility with respect to the Derby.

[Cratos]

Originally Posted by Grits

In all of its negativity, this too, is the feeling I got. But, I felt this about the entire piece. His boredom, his bitterness, outweighed any plus, even any gratitude that could have been realized through (and by) his contribution to the sport. Maybe he'll feel better after reading some of the posts that you gentlemen have written. He won't, at mine. But its OK. Obviously, this man had a good run. I hope he'll remember this.. JMO

So with you it is Dr. Roman’s personality and demeanor; and not his work that turns you off.

Therefore you probably wouldn’t have like Newton or Einstein because both of them fit the pompousness you are alleging about Roman.

[tucker6]

Originally Posted by Cratos

So with you it is Dr. Roman’s personality and demeanor; and not his work that turns you off.

Therefore you probably wouldn’t have like Newton or Einstein because both of them fit the pompousness you are alleging about Roman.

Except Roman was never and is not in the same universe as Newton or Einstein.

[dilanesp]

Originally Posted by tucker6

Except Roman was never and is not in the same universe as Newton or Einstein.

Well, I believe Einstein's theories allow for a parallel universe under certain conditions.