QUIRIN COMPUTER GENERATED SYSTEM

[SPEEDHORSE]

Anybody here who has tried WILLIAM QUIRIN'S COMPUTER GENERATED SYSTEM: Multiple Regression Formulas written in Chapter 25 of his book WINNING AT THE RACES? Are the formulas still valid today? If not what are your updated formulas. I kinda like his approach because it is not a stand alone system.

Since today's PCs can do the work today than in 1988 when I bought the book. I would like to hear first, from someone who has tried his "magic" formulas, before I start using it.

If this topic was already discussed here before, kindly direct me to the old thread. GOD BLESS !!!

[iwearpurple]

I have tried it various times. Never have I been able to show a profit.

Doesn't work.

[thaskalos]

Quirin himself admitted that it didn't work.

[Overlay]

Originally Posted by thaskalos

Quirin himself admitted that it didn't work.

Was his admission to the effect that the (admittedly limited) sample of 646 sprints and 300 routes for which he claimed a positive return (at that time) for the multiple regression formulas involved statistical practices in sampling or otherwise that would not/did not withstand a more extensive test, or was he saying that the original results themselves had been somehow deliberately inaccurate as presented?

[thaskalos]

Originally Posted by Overlay

Was his admission to the effect that the (admittedly limited) sample of 646 sprints and 300 routes for which he claimed a positive return (at that time) for the multiple regression formulas involved statistical practices in sampling or otherwise that would not/did not withstand a more extensive test, or was he saying that the original results themselves had been somehow deliberately inaccurate as presented?

He admitted that the initial positive returns attained by the multiple regression formulas could not be maintained over more extensive testing...and that nothing he could do could change that.

[Overlay]

Wouldn't that then call into question not just the multiple-regression "systems" that Quirin discussed, but all the individual factors that he identified by a single (positive) or double (negative) asterisk, which were based on similar comparatively small past samples, but which Quirin stated would retain their positive or negative character into the future, based on statistical testing of the sample results, which (according to him) indicated that those results were outside the span of normally expected deviation?

[thaskalos]

Originally Posted by Overlay

Wouldn't that then call into question not just the multiple-regression "systems" that Quirin discussed, but all the individual factors that he identified by a single (positive) or double (negative) asterisk, which were based on similar comparatively small past samples, but which Quirin stated would retain their positive or negative character into the future, based on statistical testing of the sample results, which (according to him) indicated that those results were outside the span of normally expected deviation?

Yes...it would.

[Robert Goren]

Originally Posted by Overlay

Wouldn't that then call into question not just the multiple-regression "systems" that Quirin discussed, but all the individual factors that he identified by a single (positive) or double (negative) asterisk, which were based on similar comparatively small past samples, but which Quirin stated would retain their positive or negative character into the future, based on statistical testing of the sample results, which (according to him) indicated that those results were outside the span of normally expected deviation?

It would. It maybe possible to use a multi regression system if one of the factors is the odds. That what I did back in early 70s with a very small sample by today standards. It work for about 15 years until the training methods changed. The odds (actually the log of the odds normalised for each race) was the main factor. The second leading factor was the number of starts of that year. the third was the number of starts for the previous year. the more starts the better. Money won count too and so did days since the last race. Nothing else mattered much. They must have ben reflected in the odds.
These things are hard to set up and if you going to do one , you should have someone who knows how to do it that knows nothing about horse racing helping you. That way you have fresh eyes on it. you going to have to look at data that is available that is generally the public pretty much ignores like starts was in the the 70s.
Personally I would do it again because you are looking at why the fifth place horse bets the sixth place as well why the the first place horse beats the the sixth place horse. I am not sure that same factors are in effect there.
I am doing a something now that only looks at the winner vrs the others horses in the race. I am doing in two horse match ups. It is very slow process and for health reasons I limited the amount of time I put into it. It may take me year to just to figure out often the average 2/1 beats the average 5/1 or the average 4/5 shot beats the average 10/1 if one of them is the winner. after I get a feel for what are good and reasonable numbers then I will look for thing that increases or decrease those chances by looking at other factors. I am thinking the results may look like chess ratings and the predictions operate the same way they do. I may not live long enough never finish this project. But if I do, I think I will have a solid platform for beating the ponies with a little tweaking from time to time. It will be a default system. I would still have watch out for track bias and if at the track sore horse horses. And of course wise guy trainers moves.

[ranchwest]

Multiple regression can have some significant pitfalls.

I tried the Quirin formulas years ago. One factor that I particularly didn't like was the route penalty for post position. Making a blanket assumption that a horse is doomed from an outside post is not wise.

[traynor]

Originally Posted by SPEEDHORSE

Anybody here who has tried WILLIAM QUIRIN'S COMPUTER GENERATED SYSTEM: Multiple Regression Formulas written in Chapter 25 of his book WINNING AT THE RACES? Are the formulas still valid today? If not what are your updated formulas. I kinda like his approach because it is not a stand alone system.

Since today's PCs can do the work today than in 1988 when I bought the book. I would like to hear first, from someone who has tried his "magic" formulas, before I start using it.

If this topic was already discussed here before, kindly direct me to the old thread. GOD BLESS !!!

Quirin's observations were based on a data sample of 7000 or so races. That number of races today would be considered by most as too small to generate anything but statistical noise.

It is not unusual that his formulas did not hold up when applied to other (or larger) samples. That is inherent in the fallacy of using small samples to create models, even for PhDs in Computer Science and Mathematics (who should know better, and probably do, but who may not be able to resist the easy money of writing what people want to read).

One of the things still used by a number of fairly serious bettors is the Quirin Purse Value formula. Probably because so few others use it. It is like APV, but instead of the (approximate) 0.6, 0.2, 0.1, Quirin used 0.2, 0.5, 0.8.

[Robert Goren]

Quirin made several bad choices in what he look at and the way he looked at. But remember he was limited in what he could do. It was the 80's. Not only wasn't he was able to make it work, but racing has changed a lot since then. Yet people still think his stuff with a little tweaking will work today. Just crazy!

[raybo]

Originally Posted by Robert Goren

Quirin made several bad choices in what he look at and the way he looked at. But remember he was limited in what he could do. It was the 80's. Not only wasn't he was able to make it work, but racing has changed a lot since then. Yet people still think his stuff with a little tweaking will work today. Just crazy!

The following assumes you are talking about Quirin early speed points.

I suppose that depends on how you are using Quirin. If you think a Quirin of 8 versus a Quirin of 6 means that the 8 has more early speed than the 6, then you are truly in trouble. Quirin points only help you determine "likely" early pace matchups, between horses, regarding preferred running styles, they do not tell you early speed of those horses. Just because the Quirins tell you a horse likes the lead, or likes to press the lead, doesn't mean that will happen. Other metrics need to be used to determine if that 8 has enough early speed to get or press the lead in that particular field.

[Robert Goren]

Quirin speed point are probably the only half way useful thing he ever produced. Even then they have limited value and have to be used right. A "7" or an "8" horse who isn't fast enough to get the lead is dead meat. That is about the extent of their usefulness.

[raybo]

Originally Posted by Robert Goren

Quirin speed point are probably the only half way useful thing he ever produced. Even then they have limited value and have to be used right. A "7" or an "8" horse who isn't fast enough to get the lead is dead meat. That is about the extent of their usefulness.

Shouldn't you have said: "That is the extent of their usefulness, to me"

[thaskalos]

Originally Posted by raybo

Shouldn't you have said: "That is the extent of their usefulness, to me"

You dare question the opinion of a man who has earned a profit betting horses in 45 of the last 47 years...with legitimate excuses for the other two?