Using a Class number with other figs

[Gapfire]

All feedback appreciated:

I'm thinking about adding a Mitchell style class fig to our members area, and then using it along with our other figures to create an automated odds-line.

My problem at this point, is to determine the impact value that a class number should have on the odds-line.

How do you people here at PA treat class in your analysis?

I would appreciate any responses, as I think this will give me a pretty good idea as to what type of impact value I should give to Class.

Thanks in advance,

Kevin
www.gapfire.com

[cj]

My personal opinion is the more you put class into a number, the more your line will coincide with that of the public.

[Gapfire]

My personal opinion is the more you put class into a number, the more your line will coincide with that of the public.

Thank you, I realize this. However, I don't know that ignoring class completely is the answer. I do want the odds-line to reflect real probabilities as close as possible. There must be some happy medium out there that I can incorporate into the odds-line.

[cj]

I understand, but surely you know any speed / pace numbers have a large quantity of class built in.

[Gapfire]

I understand, but surely you know any speed / pace numbers have a large quantity of class built in.

Good point. So, if you were doing and odds-line, you would just use your pace/speed numbers, and go with that? I agree with you in that pace and speed are large components of class.

[cj]

Good point. So, if you were doing and odds-line, you would just use your pace/speed numbers, and go with that? I agree with you in that pace and speed are large components of class.

The only time I use other factors is when they don't apply...surface, trainer changes, 1sters, etc.

[Greyfox]

Good point. So, if you were doing and odds-line, you would just use your pace/speed numbers, and go with that? I agree with you in that pace and speed are large components of class.

Pace and Speed are large components of class.
It is interesting to note though that even Pace "Guru" Sartin always did an earnings per start calculation in addition to the other algorithms that he was pumping out. Surely, some "Class Handicappers" on this board should have some suggestions about impact value re: Class = $$$

[jasperson]

All feedback appreciated:

I'm thinking about adding a Mitchell style class fig to our members area, and then using it along with our other figures to create an automated odds-line.

My problem at this point, is to determine the impact value that a class number should have on the odds-line.

How do you people here at PA treat class in your analysis?

I would appreciate any responses, as I think this will give me a pretty good idea as to what type of impact value I should give to Class.

Thanks in advance,

Kevin
www.gapfire.com (http://www.gapfire.com)

It depends on how you measure class. I have tried it 3 ways. Average class as defined by the average of the best of the last 3 races, IV=1.45. Best class of the last 10 races IV=1.42. Average earnings per start this year the IV=1.92 for all races. Since that is the best that is what I use.
Jack

[fmolf]

Pace and Speed are large components of class.
It is interesting to note though that even Pace "Guru" Sartin always did an earnings per start calculation in addition to the other algorithms that he was pumping out. Surely, some "Class Handicappers" on this board should have some suggestions about impact value re: Class = $$$
what might work is something i have tried to use with limited success.That is to use the earnings per start but only in races where the connections were trying.This might be tiresome and tedious because you need to discern when they were trying and when they were out for conditioning.Generally you can take all the horses races at this level and below.If a horse has shown he cannot win first off the layoff,throw those out as well.The earnings per start number will be better than what most people are using.Perhaps with the aid of a computer you could shorten the process.I am a computer illiterate.I can post here view websites and download pp's that is it....lol....good luck with your experiment.

[Gapfire]

It depends on how you measure class. I have tried it 3 ways. Average class as defined by the average of the best of the last 3 races, IV=1.45. Best class of the last 10 races IV=1.42. Average earnings per start this year the IV=1.92 for all races. Since that is the best that is what I use.
Jack

Looks like a good method that would be easy to implement.

[Gapfire]

what might work is something i have tried to use with limited success.That is to use the earnings per start but only in races where the connections were trying.This might be tiresome and tedious because you need to discern when they were trying and when they were out for conditioning.Generally you can take all the horses races at this level and below.If a horse has shown he cannot win first off the layoff,throw those out as well.The earnings per start number will be better than what most people are using.Perhaps with the aid of a computer you could shorten the process.I am a computer illiterate.I can post here view websites and download pp's that is it....lol....good luck with your experiment.
fmolf

I really like this idea, but with the multitude of races it would be impossible for me to go through each race and determine which pacelines where the connections were trying.

[GaryG]

I just don't like the idea of mucking up my figures with class values. In my handicapping I use class as a separate factor and have never tried to make an all-encompassing "power figure". I suppose there has always been a demand for one number that truly reflects a horse's chances, but I have never seen one that worked. However, I love the TPR created by our friend (?) Dick Schmidt and Tom Hambleton. No class numbers there though. Dick, thanks again. I have made a lot of money with that concept.

[fmolf]

I just don't like the idea of mucking up my figures with class values. In my handicapping I use class as a separate factor and have never tried to make an all-encompassing "power figure". I suppose there has always been a demand for one number that truly reflects a horse's chances, but I have never seen one that worked. However, I love the TPR created by our friend (?) Dick Schmidt and Tom Hambleton. No class numbers there though. Dick, thanks again. I have made a lot of money with that concept.
Has anyone tried using the equibase power ratings off of their platinum plus product?They claim to use pace ,speed and class to arrive at this number,depending on what has been winning at this track at this class level recently.My feeling about these pp,s is that they rank thefield anddo not give any par numbers .their pace figs and power ratings are rankingsof the horses in the field.

[PhantomOnTour]

Quirin advocated a technique to use when a horse runs a huge fig for his class level in a winning preformance and now steps up. Take the class of his last race and average that with the fig of his last race (par is 100 and he ran a 106...adjust his fig to a 103). Mainly applies to lower level Clm ranks as a way of 'building class into the figure' as he put it. A good way to gauge how far that winner can really move up with success. Instead of that Clm10 running a fig equal to the Clm25-30 level (106 par) he may fit better at the Clm17-20 level (103 par).

I have no testimony to its success as I haven't used it...anyone do this? A lot of the time those big fig low level winners turn out to be fool's gold even though their number is good enough for today's level.

[illinoisbred]

Quirin advocated a technique to use when a horse runs a huge fig for his class level in a winning preformance and now steps up. Take the class of his last race and average that with the fig of his last race (par is 100 and he ran a 106...adjust his fig to a 103). Mainly applies to lower level Clm ranks as a way of 'building class into the figure' as he put it. A good way to gauge how far that winner can really move up with success. Instead of that Clm10 running a fig equal to the Clm25-30 level (106 par) he may fit better at the Clm17-20 level (103 par).

I have no testimony to its success as I haven't used it...anyone do this? A lot of the time those big fig low level winners turn out to be fool's gold even though their number is good enough for today's level.
I remember reading that but no,I've never done that.Also, never done what the most popular figure seller in Chicagoland does-knock down all big runaway wins to just a 3 length winning margin. Why not leave the adjusted figure alone? That's what the horse did that day,in that race. Most of us at 1st glance could probably make a fairly accurate assessment on whether the big figure can be reproduced. I think tinkering with adjusted figure using other external factors just clouds the bigger picture. Why knock a 10,000 claimer that runs a 106 down to a 103 when most likely it won't even be able to reproduce the 103?

[Tom]

I used to use that Quirin method, and it worked very well, but nowadays, not so much. Class levels and pars have changed too much.

[GaryG]

Class levels and pars have changed too much.That is for sure. Class is not nearly the factor that it once was. More and more allowance races are being won by claimers. We are breeding cheap horses that can run only short distances.

[Cratos]

All feedback appreciated:

I'm thinking about adding a Mitchell style class fig to our members area, and then using it along with our other figures to create an automated odds-line.

My problem at this point, is to determine the impact value that a class number should have on the odds-line.

How do you people here at PA treat class in your analysis?

I would appreciate any responses, as I think this will give me a pretty good idea as to what type of impact value I should give to Class.

Thanks in advance,

Kevin
www.gapfire.com

Class is nonparametric and in horseracing it is post-race determined; it is virtually impossible to quantitate class in a continuum.

[classhandicapper]

It depends on how you measure class. I have tried it 3 ways. Average class as defined by the average of the best of the last 3 races, IV=1.45. Best class of the last 10 races IV=1.42. Average earnings per start this year the IV=1.92 for all races. Since that is the best that is what I use.
Jack

Thanks for those numbers.

The major problem with AEPS is that some purses are inflated. The other is that it probably outperforms the other measurements because it incorporates consistency. Horses that are more consistent earn more money per start and do better even when they are of equal class.

If you added a consistency measure to the others, they would probably do better.

If I was going to try to create a formula for my subjective thinking I would create a combination of a consistency method that weighs win, places, and shows appropriately + overall level of competition + best class recently.

My own thinking is a little more complex than that and changes depending on circumstances, but that's the gist of it.

[gm10]

All feedback appreciated:

I'm thinking about adding a Mitchell style class fig to our members area, and then using it along with our other figures to create an automated odds-line.

My problem at this point, is to determine the impact value that a class number should have on the odds-line.

How do you people here at PA treat class in your analysis?

I would appreciate any responses, as I think this will give me a pretty good idea as to what type of impact value I should give to Class.

Thanks in advance,

Kevin
www.gapfire.com

I use in two ways:

* take an 'average' of recent class numbers for races where the horse managed to beat at least half the field. This can be an input factor for probability generating models.
* for today's race, calculate the probability a horse will run a speed number at least as high as the class number (assuming you have some statistical distribution for your horse's speed figures, and speed and class numbers are on the same scale). Normalize those probabilities, or use them as input variables.

[Greyfox]

Class is nonparametric and in horseracing it is post-race determined; it is virtually impossible to quantitate class in a continuum.

1. Not according to the late Dick Mitchell, who was using it to create an odds line.

2. Thorndike once said:
"If something exists, it exists in quantity. Anything that exists in quantity can be measured."

While a "continuum" might not be possible, certainly a weighting factor can be integrated into any algorithm. If that algorithm works better than whatever you're using now, it's a keeper, independent of the supposed impossibility of measuring "Class."

[skate]

a key to class, not mentioned, is the track you play.


To consider Class at HP, is not even close to considering Class at Remington.

Even with Md Sp. Wts, who have not run a race, you should consider its possible class (which i think everyone can see this), but not so at lesser class tracks.

you would only know this if you play both tracks.;)

Not to mention what an exagerated Purse might do.

[Trotman]

Pace time has been the priority with my handicapping a race. That being said a horse or horses that can set the pace or overcome it are todays class. Every race is an independent event where assigning a number to class IMO is futile and clouds any chance of betting a winner.

[Cratos]

1. Not according to the late Dick Mitchell, who was using it to create an odds line.

2. Thorndike once said:
"If something exists, it exists in quantity. Anything that exists in quantity can be measured."

While a "continuum" might not be possible, certainly a weighting factor can be integrated into any algorithm. If that algorithm works better than whatever you're using now, it's a keeper, independent of the supposed impossibility of measuring "Class."

I respect the work of Thorndike in the field of psychological measurement, but using his theory to quantitate a boundless variable is useless.

For example a horse wins a claiming race, we can say in the post-race that horse is a $XXX claimer. However if it is entered in an allowance in its next race and wins, it becomes an allowance horse and somewhere in its career it will find its class level; but its class can never be pre-race determined

[thaskalos]

I respect the work of Thorndike in the field of psychological measurement, but using his theory to quantitate a boundless variable is useless.

For example a horse wins a claiming race, we can say in the post-race that horse is a $XXX claimer. However if it is entered in an allowance in its next race and wins, it becomes an allowance horse and somewhere in its career it will find its class level; but its class can never be pre-race determinedJust because a claimer beat a so-called allowance field, that does not make the horse a legitimate allowance horse. It depends on the allowance race that it prevailed in.

The allowance race could have been a mix of other claimers and off-form allowance horses. These class structures are not as precise as we would like them to be.

[Tom]

Class can be pre-determined and is everyday. You are also wrong a lot. It is not a number, like a pace number.

[thaskalos]

Pace time has been the priority with my handicapping a race. That being said a horse or horses that can set the pace or overcome it are todays class. Every race is an independent event where assigning a number to class IMO is futile and clouds any chance of betting a winner.I agree that assigning a number to class is futile, because I have been trying to create one for many years...with dubious results.

But I disagree that pace can be a determinant factor in defining class, even in sprints. Every day, in every track in the country, many seemingly faster horses (according to their superior speed and pace figures) are raised in class and lose badly to "classier" horses that outwardly appear to be much slower than them.

It behooves the player to be very leary of outclassed horses, even if they have a clear pace edge in the race, because the pace advantage very often fails to materialize when the horses rise in class.

[Charlie D]

Just because a claimer beat a so-called allowance field, that does not make the horse a legitimate allowance horse. It depends on the allowance race that it prevailed in.

The allowance race could have been a mix of other claimers and off-form allowance horses. These class structures are not as precise as we would like them to be.


I think you describe nicely thaskalos, the major problem with the so called "Class structure" measurement.

[cj]

Just because a claimer beat a so-called allowance field, that does not make the horse a legitimate allowance horse. It depends on the allowance race that it prevailed in.

The allowance race could have been a mix of other claimers and off-form allowance horses. These class structures are not as precise as we would like them to be.

Exactly...the same thing could be said about G1 Stakes races.

[Charlie D]

Exactly...the same thing can be said about G1 Stakes races.

FTFY :)

[thaskalos]

FTFY :)Charlie...is he talking about Zenyatta again? :)

[Charlie D]

Charlie...is he talking about Zenyatta again? :)

Only CJ can answer that, but her G1's can be used as examples can't they??


Lets try not to turn this interesting thread into another boring Zen debate though :)

[thaskalos]

Only CJ can answer that, but her G1's can be used as examples can't they??


Lets try not to turn this interesting thread into another boring Zen debate though :)You won't hear a peep out of me...I'm not in a debating mood.

[PhantomOnTour]

Pace time has been the priority with my handicapping a race. That being said a horse or horses that can set the pace or overcome it are todays class. Every race is an independent event where assigning a number to class IMO is futile and clouds any chance of betting a winner.
That is how every horse race ever run has been won, by setting and/or overcoming the pace...with the exception of DQs and spills :)

[Charlie D]

That is how every horse race ever run has been won, by setting and/or overcoming the pace...with the exception of DQs and spills :)


Correct. but don't tell everyone Phantom :)

[Overlay]

The major problem with AEPS is that some purses are inflated.

That's what they said when purses started being inflated with state-bred money in the 1980's, and yet, fifteen years later, Nunamaker found that average earnings-per-start was still just as effective with respect to impact values as Quirin had found it to be.

Also, although speed figures do reflect class, speed figures were around at the time of Quirin's research, and his multiple-regression formulas included a class component based on AEPS (in addition to a speed component) because of the consistency aspect that AEPS encompassed.

To me, earnings is still the way to cut through the fog of conflicting or inconsistent class-level labels or designations. It's in the interest of better horses to race for the biggest purses for which they are eligible, and those horses will, in turn, exhibit more consistent performance under those conditions.

But, again, that's only one weighted part of the overall handicapping picture, including (especially) wagering value.

[thaskalos]

That's what they said when purses started being inflated with state-bred money in the 1980's, and yet, fifteen years later, Nunamaker found that average earnings-per-start was still just as effective as Quirin had found it to be.

Also, although speed figures do reflect class, speed figures were around at the time of Quirin's research, and his multiple-regression formulas included a class component based on AEPS (in addition to a speed component) because of the consistency aspect that AEPS encompassed.

To me, earnings is still the way to cut through the fog of conflicting or inconsistent class-level labels or designations. It's in the interest of better horses to race for the biggest purses for which they are eligible, and those horses will, in turn, exhibt more consistent performance under those conditions.

But, again, that's only one weighted part of the overall handicapping picture, including (especially) wagering value.Hi Overlay,

I read your work several years ago (very impressive!), and I'm wondering if you have continued your research in an attempt to refine your methods even more.

I have been trying for a long time to quantify class with a number - or numbers - and have not succeeded to my satisfaction. I have also been dissatisfied with commercially available class figures.

I have tried factoring earnings into the equation, but the fact that earnings do not differentiate between recent class and past class makes me hesitant about accepting them as valid indicators.

I know that they are still effective as a line-making component...but would you still recommend earnings in a "non" line-making, comprehensive, power rating-type process?

[Overlay]

Hi Overlay,

I read your work several years ago (very impressive!), and I'm wondering if you have continued your research in an attempt to refine your methods even more.

I have tried factoring earnings into the equation, but the fact that earnings do not differentiate between recent class and past class makes me hesitant about accepting them as valid indicators.

I know that they are still effective as a line-making component...but would you still recommend earnings in a "non" line-making, comprehensive, power rating-type process?

I understand your point about "present class" versus "past class", but I would consider class by itself (without going too far back into ancient history)(as you would know), and then rely on more recent condition and speed elements to provide an indication of the horse's potential to perform to whatever class level it had shown that it might be capable of attaining. The consideration of consistency as part of class also aids in this.

As you would also know, my handicapping process produces an overall value for each entrant that incorporates all of the elements that I look at. That value is what I would consider if I were only interested in a single figure for each horse from the standpoint of which horse is likeliest to win. Line-making just adds the additional step of comparing the respective values of all the horses in the race to develop an estimate of winning probability for each.

Thanks for the favorable mention! :)

[Sericm]

Class is the Horse who get's to the finish line first no matter what the race.
To the swift goes the victory.

[thaskalos]

Class is the Horse who get's to the finish line first no matter what the race.
To the swift goes the victory.Wow...nothing to it!

No wonder this game is so easy to beat.

[Gapfire]

I have been trying for a long time to quantify class with a number - or numbers - and have not succeeded to my satisfaction. I have also been dissatisfied with commercially available class figures.



This is precisely why I started this thread. Great responses, and ideas so far. :ThmbUp:

[ezpace]

Trotman msg#23 and Cj posts are A=1

I'm not big into averaging speed/pace numbers for their

purpose or class normally but i make a number that avgs

best 2 final times at sur/dist and then require that

number to be beaten by my proprietary 3rd fraction #

either by the E type (if it sets up that way) or the P,or S horse

[Trotman]

Class has always been to me anyway TODAY'S CLASS as I feel trying to quantify by a number or whatever on BACK CLASS is like shooting yourself in the foot.

[Cratos]

Just because a claimer beat a so-called allowance field, that does not make the horse a legitimate allowance horse. It depends on the allowance race that it prevailed in.

The allowance race could have been a mix of other claimers and off-form allowance horses. These class structures are not as precise as we would like them to be.

You are absolutely correct