What are the best track variants ?

[cosmicway]

The computation of track variant is central to horseracing prediction theory.
We see a variety of methods advertised in the web but which one is the best ?

The standard method, as I gathered is this:

1 - Have available a set of class pars (average winner times) for all the types of races in the track.
2 - For each race subtract the class par from the recorded winner's time.
3 - The difference is the race variant.

Is this correct ?

What is the criterion the makers of various track variant systems seek to fulfill ?
The ultimate criterion is of course some sort of p.d.f. (probability distribution function) but how can they tell in advance if what they are doing is correct ?

[Jed]

my vote is : That is correct.

[Tom]

No matter what method you use, class pars, projeciton, or anything elses out there. remeber, your result is ALWAYS a guess and NEVER reality. Reality is what tim ethe race went in. An adjusted time is only a guess as to what it SHOULD have gone in.
Some guesses are better than others, and some people use raw times and win. And add in that there is no such thing as an accurate lengths behind call for anyone but the leader, and the guess is more and more a guess.
I never cease to be amused by the accuracy some people go to to get times down to 1/100s, but then use BL that might by 2-5 lengths off to start with.
But if they win money with it, my opinion doesn' t matter. Bottom line is bottom line.

[thoroughbred]

[QUOTE=cosmicway]The computation of track variant is central to horseracing prediction theory.
We see a variety of methods advertised in the web but which one is the best ?

The standard method, as I gathered is this:

1 - Have available a set of class pars (average winner times) for all the types of races in the track.
2 - For each race subtract the class par from the recorded winner's time.
3 - The difference is the race variant.

Is this correct ?

--------------------------------------------------------------
I do not believe you are completely correct. Track variant is supposed to represent the variation of the TRACK from its par value. The winning horse may have been either a superior horse, or a not so great horse and the difference in time from the par value may have more to do with the particular HORSE than the track.

[cosmicway]

cosmicway wrote
-----------------
The computation of track variant is central to horseracing prediction theory.
We see a variety of methods advertised in the web but which one is the best ?

The standard method, as I gathered is this:

1 - Have available a set of class pars (average winner times) for all the types of races in the track.
2 - For each race subtract the class par from the recorded winner's time.
3 - The difference is the race variant.

Is this correct ?

thoroughbred wrote
-------------------
I do not believe you are completely correct. Track variant is supposed to represent the variation of the TRACK from its par value. The winning horse may have been either a superior horse, or a not so great horse and the difference in time from the par value may have more to do with the particular HORSE than the track.
__________________


I said this is what I understood.
Do you know a better definition ?

I believe what is called for is to work out some sort of race to race equilibration.
I read somewhere that if you take human athletes to compete in different cities (Athens-Munich-Tokyo a.s.o) it's going to be the same thing - unless they encounter extremes of weather.
But with horses adjustments are nearly always necessary. It is a combination of ground conditions - pace injected and after all you have two athletes working together (horse and jockey) and the situation becomes complex.

I believe in race variants as opposed to daily variants.
A recent sprint race is what prompted me to start working on the issue.
The best sprint horse we have here (Kathmor owned by a company called "Black Diamond") ran 5 furlongs at 57'' and won - a dismal time figure but the opposition was pretty inferior and infact not existent (owing to a number of early morning scratches). So the race variant has to be some -2 in my reckoning but the books said 0 (normal !) and I don't know on what sort of model they based their figure. Practically the horse finished eased down.
It's ludicrous and Kathmor in his next race went on to score his usual 55'' and won again, against some more realistic opposition this time.

There is a fly in the ointment I said.
I worked out a scheme of track variants based on the differences between successive races of the same horse and asked for the values that minimize the mean difference (to be produced).
This is different from a class par based method but it apparently does not work well when you try to predict future winners, although at first it looked brilliant.

Everything is of course statistical.
If you can predict the future winner with an average success rate 35% and I am held at 32% then you win. Nobody looks forward to predict all races.

Anyway I am interested in buying a model to apply it, not to bet in U.S.A. (that I can't do under the current restrictions as it were).
Which one is the best and the algorithm (or procedure) is visible ?

[toetoe]

No, sir. If my average mutuel is as little as 11/10 of yours, I'm ahead of you, even 3 more of your horses win per 100 races.

[cosmicway]

toetoe says:
> No, sir. If my average mutuel is as little as 11/10 of yours, I'm ahead of you, even 3 more of your horses win per 100 races.

That's what I 'm saying.