Fractional Variant Theorem

[michiken]

Suppose that the PAR times (in hundreths of seconds) for a theoretical mile race are given as:

2f = 24.00
4f = 48.00
6f = 72.00 (1:12)
8f = 96.00 (1:36)

Now imagine a race in the middle of the February winter at you favorite cheap night track. A mile race is run in:

2f = 25.00
4f = 51.00
6f = 76.00 (1:16)
8f = 104.00 (1:44)

Based on a Final Time variant formulae, this race would be deemed (104 - 96) = 8 seconds slow.

Now consider creating a Fractional Variants based on the par times:

V1 = 1.00 (25.00 - 24.00)
V2 = 3.00 (51.00 - 48.00)
V3 = 4.00 (76.00 - 72.00)
V4 = 8.00 (104.00 - 96.00)

Fractional Combined Variant = 16.00 seconds slow.

THE DIFFERENCE between the tradional and fractional variants > 8.00 seconds slower.

(This type of variant could also apply to faster than par races).

I was just curious if anyone has ventured into this type of variant making? or may apply an adjustment to each horses pace figure?

I would appreciate your comments.

[ryesteve]

The reason you're getting a different answer is because you're using the difference between the first quarters three extra times, the difference between the second quarters two extra times, and the difference between the third quarters one extra time. I'm not sure why you'd want to weight a final time variant that way, but the idea of using fractional pars and variants to make pace figs certainly makes sense.

[Light]

I believe you are aware that 2 pace points = 1 speed point.Therefore your fractional times should equal your SR,which yours does.

[cj]

I don't use your method, but I've made pace pars and variants for years now. There is certainly value to doing this. However, I would not ever recommend using a fraction of the speed variant.

[Dave Schwartz]

IMHO, you are making this way too complicated. Please permit me to offer a better solution.

First, whatever you deem your variant to be for the final time is THE VARIANT for the race.

You adjust your pace times via the relationship between the original pace times and the original final time.

Thus, in your example, the race times were:
2f = 25.00
4f = 51.00
6f = 76.00 (1:16)
8f = 104.00 (1:44)

If you have chosen to make the final time 8 seconds faster, and call it 1:36.00, then you wind up with:

2f= 23.07 computed by: 25.00 / 104.00 x 96.00
4f= 47.07 computed by: 51.00 / 104.00 x 96.00
6f= 70.15 computed by 76.00 / 104.00 x 96.00


V1 = 1.93
V2 = 3.96
V3 = 5.85
V4 = 8.00

[Dr.SwineSmeller]

Did Andrewfus Beyer help dream this newfangled algebraic equation up too? His other method is lousy as evidenced by his Belmont predictions past.

I don't think Andrewfus has picked the Belmont Stakes winner in the entire last decade? Albert Einstien would be so proud of Drewfus for his Jethro Bodeen style of mathematical sipherin'... Ought and ought makes ought..

Dr. SwineSmeller

[46zilzal]

I tried this in the early 80's and there is some merit to the idea.

[Tom]

I used to make pace variants for three calls, and a lot of times a lone front runner who ran exceptionally fast early on and then died threw the early variants out of wack.

[cj]

Did Andrewfus Beyer help dream this newfangled algebraic equation up too? His other method is lousy as evidenced by his Belmont predictions past.

I don't think Andrewfus has picked the Belmont Stakes winner in the entire last decade? Albert Einstien would be so proud of Drewfus for his Jethro Bodeen style of mathematical sipherin'... Ought and ought makes ought..

Dr. SwineSmeller

Another Beyer hater. This time you drag his name into a thread that has ZERO to do with him. Please try to stay on topic.

The Belmont predictions thing is assinine. The Belmont is not a race that lends itself to speed figure handicapping.

[cj]

I used to make pace variants for three calls, and a lot of times a lone front runner who ran exceptionally fast early on and then died threw the early variants out of wack.

Exactly, which is why I use a hopefully still unique method to make them.

[Tom]

Yeah, CJ.....that's why I say is USED to make variants.;)

[ezpace]

winners line all
the way through,
right cj.
i know iknow
S the F up ;) ;)

okok ;)

[Dr.SwineSmeller]

I tried this in the early 80's and there is some merit to the idea.

So what I hear you saying is, "It weren't that good a sipherin' method or I'd still be doing it..."

Dr. SwineSmeller

[46zilzal]

So what I hear you saying is, "It weren't that good a sipherin' method or I'd still be doing it..."


No it was too much work (had to manually do it then) and I found a method that extracts if from a data source via a computer program. Still works fine.

[Dr.SwineSmeller]

Another Beyer hater. This time you drag his name into a thread that has ZERO to do with him. Please try to stay on topic.

The Belmont predictions thing is assinine. The Belmont is not a race that lends itself to speed figure handicapping.

I don't hate Beyer. I just have no pity for a self-appointed handicapper who GUARANTEES one horse can outrun ten other horses in ANY horse race.

Public handicappers that are in the position to sway the general public's wagers, don't guarantee anything in a horse race if they a true "Handicapper". To me, when you fail on a guarantee, you lose your credibility in any business, including, and especially horse racing.

In fact, "guarantees" go directly against Webster's NWC Dictionary definition of what a handicapper is: [quote] handicapper> "a person, as a sports writer, who TRIES[not same as guarantee] to predict the winners in horse races on the basis of past records, track conditions, ect..."

Andrew should know the proper definition of his self-appointed title before he goes self-appointing himself.

Did you ever hear Randy Moss guarantee a win? Nah, he's a real handicapper, no guarantees...

By the time I finish with Drewfus, he may never make another guarantee in a horse race as long as he lives. Maybe he will, there is a sucker born every day..

Note my signature below CJ, it actually has meaning...With Andrew's guarantee, he becomes that metaphorical pig that turned around and messed in the hog trough at feedin' time.

Ok CJ, I'll take it to another thread.

(Unless someone else wants to talk about it some more in this one...).

Dr. SwineSmeller

[sjk]

I think that if you make pace figures you must look at sectional track variants. If you do not the days when the wind blows or when one section of the track has been maintained differently than the rest will mess with your bankroll.

I tend to disagree with Dave's simplified calculation. The effect of the day's variant on the fractional times is back-end loaded. I would adjust the fractional times to something more like:

24.3, 48.7, 71.2

[xtb]

By the time I finish with Drewfus, he may never make another guarantee in a horse race as long as he lives.


He must be shaking right now.

[Lose The Juice]

People have been doing this on the harness side for ages. It used to work, sometimes wildly well. But the problem, as is also the problem on the t-bred side, is that horse drugging has wildly increased race-to-race variability, to the point at which recent performance becomes almost a secondary issue.

[cj]

People have been doing this on the harness side for ages. It used to work, sometimes wildly well. But the problem, as is also the problem on the t-bred side, is that horse drugging has wildly increased race-to-race variability, to the point at which recent performance becomes almost a secondary issue.

Ridiculous.

[HUSKER55]

I don't think you can take a track variant and say, for example, "How fast is slow", or "How fast is fast".

When I talked to the girl at DRF my interpretation, which she agreed to, was that their variant gives the average speed rating for winners than day. Therefore, if the horse in question has a speed rate of 88 and the variant is 20 then this horse ran above average that day.

Would someone please explain for me how you can read anything more than that into a track variant.

BTW, be kind, I am still learning.

Thanks

husker55

:)

[PaceAdvantage]

A note to all moderators:

Please, FEEL FREE to use your discretion and DELETE any replies that are clearly off topic, or meant to take any thread clearly off topic. Dr. Swine's initial reply in this thread is a perfect example.

The quality of discussion is impacted severely by tangents such as these, and will do nothing but dissuade folks from starting meaningful discussions on handicapping in the future.

Thanks.

[Dr.SwineSmeller]

I'm out. It was fun while it lasted. Happy trails...

Dr. SwineSmeller

[OTM Al]

Did you know that the term "bite the dust" was used in the Illiad? Some things never go out of style....but back to topic

My question here would be if you are looking for winners, why would you set a pace variant on a random lead horse? One would think you would want the variant on the winner as that is the profile you are looking for. Even imperfectly adjusted from the time at the call, those miscalculations should even out over several trials. Been thinking about trying this for a while, but haven't reached motivation level yet. Anyone else tried this or is there a fundemental flaw in such an approach I am missing?

[ryesteve]

is there a fundemental flaw in such an approach I am missing?I don't know if this is a "flaw" or not... it could be a desired result; but by using the winning horse as the benchmark, the variant will not just be measuring the relative speed of the track, it'll also be measuring how speed-favoring the track was that day.

For example, if the inherent speed of the track was the same on saturday as it was on wednesday, but for whatever reason, most of the wednesday winners were coming from off the pace, while most of the saturday winners were going wire to wire, you'll end up with wednesday's variant being much higher. But, that could be good, because you'll end up inflating the pace ratings of the horses who ran against the bias... so I guess as long as you recognize that you're deriving "pace/bias" ratings and not just "pace" ratings, maybe it's all good.

[Premier Turf Club]

JMHO, but when I made figures, years ago I never tied final time variants and pace variants together. Sometimes there is a relationship between the two, sometimes not. Run-ups change, on days with a strong wind the horses are plus wind for part of the two turn race and minus wind for another part and yet the net effect might be zero, crazed frontrunner opens 5 and breaks the beam with a ridiculous pace figure in a race that falls apart, etc.

Long winded answer (pardon the pun) but I found this enough of the time, especially making figures for AQU inner that I couldn't use a fractional set rule. I might start there just to get a rough idea, but many times ended up with very different final results.

[delayjf]

A while back a poster (MVMCKEE) posted a alternate method of creating fractional variants that to me makes the most sense.

He began by asking the question, when final times differ (say by one second) how is that difference distributed among the fractional times. The percentages he came up with for six furlongs after looking at thousands of races were:
6 Furlongs:

1st fraction: 17%

2nd fraction: 33%

3rd fraction: 50%

In this case you would take the final time variant and apply 17% to the first fraction - 33% to the second etc. The way he explained it made perfect sense. As the race progresses, the fractional times run by horses are affected by two factors 1) the speed of the track 2) the increasing fatigue on the horse cause by running the previous fractions. Which is why the third fraction gets the biggest adjustment – it’s the fraction that is the most effected.

I cannot tell you if the above relationship / percentages are accurate for your track, but here is how you can determine what those fractional relationships are at any track.

Purchase an accurate set of Pars that include fractions times

Take the fractional times associated for a specific final time and compare them with the fractional times of a final time that is one second faster. Then calculate the % difference in the fractional times. You can do this several times for each distance using different final times to get an avg sprint and route variant distribution. If you feel your sample is too small use .5 second differences in final times.

[Greyfox]

Dick Mitchell set out a formula years ago for adjusting the projected pace of today's race using the track variant from the representative race. As someone has borrowed my Mitchell books, I can't provide it here. If memory serves me correct, the relationship was curvilinear not linear between pace and final time. He spoke briefly about using the square root of the variant to make slight adjustments to second call times.
At any rate, my own private program makes those slight adjustments, but trying to trace them on a spread sheet is no easy feat.
Someone with a Mitchell book might throw more light on this topic.

P.S. At some cheap tracks where the variant is over 30 using almost any technique throws the predictions well off line from what reality turns out to be.
So for practical reasons, on my spread sheet 30 is the highest variant that I will give a horse in making adjustments to time.

[Cratos]

Suppose that the PAR times (in hundreths of seconds) for a theoretical mile race are given as:

2f = 24.00
4f = 48.00
6f = 72.00 (1:12)
8f = 96.00 (1:36)

Now imagine a race in the middle of the February winter at you favorite cheap night track. A mile race is run in:

2f = 25.00
4f = 51.00
6f = 76.00 (1:16)
8f = 104.00 (1:44)

Based on a Final Time variant formulae, this race would be deemed (104 - 96) = 8 seconds slow.

Now consider creating a Fractional Variants based on the par times:

V1 = 1.00 (25.00 - 24.00)
V2 = 3.00 (51.00 - 48.00)
V3 = 4.00 (76.00 - 72.00)
V4 = 8.00 (104.00 - 96.00)

Fractional Combined Variant = 16.00 seconds slow.

THE DIFFERENCE between the tradional and fractional variants > 8.00 seconds slower.

(This type of variant could also apply to faster than par races).

I was just curious if anyone has ventured into this type of variant making? or may apply an adjustment to each horses pace figure?

I would appreciate your comments.

A race curve is a downward sloping curve which is non-linear and cannot be fitted by your assumptions.

[Dr.SwineSmeller]

[QUOTE=OTM Al]Did you know that the term "bite the dust" was used in the Illiad? Some things never go out of style....but back to topic

No I didn't know that. But thanks for the useful info Homer. I'm sure it'll come in handy someday when I regress 3000 years.

Dr. SwineSmeller

[TrifectaMike]

A race curve is a downward sloping curve which is non-linear and cannot be fitted by your assumptions.

More likely a combination of exponential growth and decay functions.

Mike

[michiken]

I used this variant method to modest success at the old Detroit Race course.

I brought up this question because it was the very first method that I used to make my figures. Of course this was years ago when I actually had time to do so.

Thanks for the interesting feedback. I will play with some of these methods when I have free time.

[delayjf]

The percentages he came up with for six furlongs after looking at thousands of races were:
6 Furlongs:

1st fraction: 17%

2nd fraction: 33%

3rd fraction: 50%

To add a touch of validity to the above, no less of a handicapper - Tom Brohamer also advocating appling 50% of the final time variant to the second call in sprints. The only time any adjustment was made to the first call was if the variant was big enough that it required the adjustment - I don't recall how big the final variant had to been before a 1st call adjustment was warrented.

[cj]

To add a touch of validity to the above, no less of a handicapper - Tom Brohamer also advocating appling 50% of the final time variant to the second call in sprints. The only time any adjustment was made to the first call was if the variant was big enough that it required the adjustment - I don't recall how big the final variant had to been before a 1st call adjustment was warrented.

Over a large number of races, it will "average" out. But there will be lots of errors in individual races.

[delayjf]

CJ,

I’ve no doubt you are correct.

There is one thing I need to correct from my earlier statement about. The percentages listed (17– 33 – 50) was based on win pars (the avg fractional / final times of the winner).

Looking at win pars, it occurred to me that one could refine the above percentages further by sorting Win Pars based on the running style of the horse. In other words a front runner’s fractional percentages break down would be different than a closers.

[thruncy]

To all variant makers: There is one; but trying to express it numerically may be a waste of time better spent on other handicapping factors.

[cj]

To all variant makers: There is one; but trying to express it numerically may be a waste of time better spent on other handicapping factors.

Anything people think isn't worth doing is usually time well spent in this game.

[sjk]

If you are willing to let the computer do the work the time is no big deal.

Even better if the process you would do by hand would take so much time that a person would never expend the energy to keep up it is still no big deal for the computer.

[raybo]

I'm resurrecting this thread as I have been doing some thinking along these lines of late and would like some input from the pace gurus here.

Let's suppose that one has a track par chart that includes par times for all the calls and the final par times for all distances run at every track in the US.

If one were to have the computer do a "lookup" function for today's track/distance/surface and compared that to every past race, for each horse in the race, adjusting the PP races' fractional times and final times (the horse's actual fractional and final times, not the leaders or winner's times) by the percentage of difference between today's track fractional/final time pars and the PP races' fractional/final times (the horse's actual times, not the leader's or the winner's), would this method lend itself to a better means of analysis of what each PP represented, and would it provide a decent means of comparing each horse to all the other horses in the race, as far as pace abilities and the resulting final times abilities?

I'm sure someone here has already studied this method and either found it viable or not. CJ and others who produce their own pace figures probably have been through this at some time in the past.

Please don't just say "It won't work." but instead, offer some sort of intelligent basis for your opinion. Thanks, in advance.

[gm10]

A while back a poster (MVMCKEE) posted a alternate method of creating fractional variants that to me makes the most sense.

He began by asking the question, when final times differ (say by one second) how is that difference distributed among the fractional times. The percentages he came up with for six furlongs after looking at thousands of races were:
6 Furlongs:

1st fraction: 17%

2nd fraction: 33%

3rd fraction: 50%

In this case you would take the final time variant and apply 17% to the first fraction - 33% to the second etc. The way he explained it made perfect sense. As the race progresses, the fractional times run by horses are affected by two factors 1) the speed of the track 2) the increasing fatigue on the horse cause by running the previous fractions. Which is why the third fraction gets the biggest adjustment – it’s the fraction that is the most effected.

I cannot tell you if the above relationship / percentages are accurate for your track, but here is how you can determine what those fractional relationships are at any track.

Purchase an accurate set of Pars that include fractions times

Take the fractional times associated for a specific final time and compare them with the fractional times of a final time that is one second faster. Then calculate the % difference in the fractional times. You can do this several times for each distance using different final times to get an avg sprint and route variant distribution. If you feel your sample is too small use .5 second differences in final times.

very good post
I had always assumed 1/3 for each fraction

[sjk]

Raybo,

What you suggest is what I think many think handicappers do.

One change I would recommend is to adjust the pp lines according to variants (for the sectional components of the race as well as the final time) for the day/surface on which the lines were run before comparing these adjusted times with the pars.

[gm10]

I'm resurrecting this thread as I have been doing some thinking along these lines of late and would like some input from the pace gurus here.

Let's suppose that one has a track par chart that includes par times for all the calls and the final par times for all distances run at every track in the US.

If one were to have the computer do a "lookup" function for today's track/distance/surface and compared that to every past race, for each horse in the race, adjusting the PP races' fractional times and final times (the horse's actual fractional and final times, not the leaders or winner's times) by the percentage of difference between today's track fractional/final time pars and the PP races' fractional/final times (the horse's actual times, not the leader's or the winner's), would this method lend itself to a better means of analysis of what each PP represented, and would it provide a decent means of comparing each horse to all the other horses in the race, as far as pace abilities and the resulting final times abilities?

I'm sure someone here has already studied this method and either found it viable or not. CJ and others who produce their own pace figures probably have been through this at some time in the past.

Please don't just say "It won't work." but instead, offer some sort of intelligent basis for your opinion. Thanks, in advance.

I won't say much about it apart from that this is a very valuable approach.

[gm10]

A while back a poster (MVMCKEE) posted a alternate method of creating fractional variants that to me makes the most sense.

He began by asking the question, when final times differ (say by one second) how is that difference distributed among the fractional times. The percentages he came up with for six furlongs after looking at thousands of races were:
6 Furlongs:

1st fraction: 17%

2nd fraction: 33%

3rd fraction: 50%

In this case you would take the final time variant and apply 17% to the first fraction - 33% to the second etc. The way he explained it made perfect sense. As the race progresses, the fractional times run by horses are affected by two factors 1) the speed of the track 2) the increasing fatigue on the horse cause by running the previous fractions. Which is why the third fraction gets the biggest adjustment – it’s the fraction that is the most effected.

I cannot tell you if the above relationship / percentages are accurate for your track, but here is how you can determine what those fractional relationships are at any track.

Purchase an accurate set of Pars that include fractions times

Take the fractional times associated for a specific final time and compare them with the fractional times of a final time that is one second faster. Then calculate the % difference in the fractional times. You can do this several times for each distance using different final times to get an avg sprint and route variant distribution. If you feel your sample is too small use .5 second differences in final times.

weird
I just had a look at 6D on the Belmont dirt
when I look at all races, I get similar numbers to yours
when I only take the races on a Fast track, I get 30% 40% 30%

[thruncy]

:bang: Are like religions...there are many true beleivers; but they can't express their beliefs with accuracy and objectivity. Ultimately, thr "believers" can only achieve subjectivity...and then there's those that express "par" times in hundredths of seconds. This is called "shobiz.":D

[Tom]

And then there are those who cash ticket after ticket using the ones they make.

[thruncy]

Look up the last 3 races run by this horse--2 @ DMR & 1 @ FPX-- and ask yourself if you still can make sense out of your "horse specific" variant prospects. I don't think so. In addition, horses are still wrapped up long before the finish line for varying reasons.

[Tom]

Horses fall down, too - it has no effect on the variant.

[delayjf]

In the e-mail above, I mistaken said that one could use par times to determine what the fractional relationships were, but that was wrong. the 17-33-50 relationship was based on win pars (avg fraction times of the race winner) not the race pars which could reflect the efforts of three different horses.

[delayjf]

when I only take the races on a Fast track, I get 30% 40% 30%
Are front runners winning a majority of the races? The above might indicate some type of speed bias.

[gm10]

Look up the last 3 races run by this horse--2 @ DMR & 1 @ FPX-- and ask yourself if you still can make sense out of your "horse specific" variant prospects. I don't think so. In addition, horses are still wrapped up long before the finish line for varying reasons.

that's the problem with polytrack and especially turf races
DTV's are often more the consequence of how the races were run that day than of how fast the track was

[cj]

that's the problem with polytrack and especially turf races
DTV's are often more the consequence of how the races were run that day than of how fast the track was

Unless, that is, the variant maker is aware of that issue and takes it into account.

[gm10]

Unless, that is, the variant maker is aware of that issue and takes it into account.

true but I do them by computer
I've tried (automated) finding ways around it but no luck so far

[raybo]

Raybo,

What you suggest is what I think many think handicappers do.

One change I would recommend is to adjust the pp lines according to variants (for the sectional components of the race as well as the final time) for the day/surface on which the lines were run before comparing these adjusted times with the pars.

I'd thought of that but am just in the initial stages of this line of thought. If variants are to be used for the fractional times as well as the final times, the question remains, how would you go about distributing the variant among the fractional times?

Would it be ok to use the distribution suggested earlier in the thread (17%, 33%, 50%)? This approach would seem to be too simplistic to me, surely every track and each distance would not have the same distribution for each fraction. On heavier tracks and at longer distances, logic would say, the track variant would affect the later fractions more than the early ones.

The following is "right off the top of my head" so bear with me.

How about finding the percentage of the "fraction vs the sum of the fractions" and then use the same percentage of the final time variant for that particular fraction?

ie; Let's assume that a race was run with the fractional times of 22 46 72. If you add the 3 fractional times together that would be 22+46+72=140, then divide the 1st fraction by this sum to get: 22/140=.1571 or about 16%, then the 2nd fraction would be: 46/140=.3285 or about 33%, and finally: 72/140=.5142 or about 51%. So you would have 16%, 33% and 51% (=100% of the variant) rather than 17%, 33%, and 50% (100% of the variant, also) as was suggested in the earlier post. I would think that this might be a more accurate way of distributing the variant from track to track and distance to distance.

What do you think?

PS. If someone is using this method, I apologize for putting it "out there".

[cj]

I'd thought of that but am just in the initial stages of this line of thought. If variants are to be used for the fractional times as well as the final times, the question remains, how would you go about distributing the variant among the fractional times?

Would it be ok to use the distribution suggested earlier in the thread (17%, 33%, 50%)? This approach would seem to be too simplistic to me, surely every track and each distance would not have the same distribution for each fraction. On heavier tracks and at longer distances, logic would say, the track variant would affect the later fractions more than the early ones.

The following is "right off the top of my head" so bear with me.

How about finding the percentage of the "fraction vs the sum of the fractions" and then use the same percentage of the final time variant for that particular fraction?

ie; Let's assume that a race was run with the fractional times of 22 46 72. If you add the 3 fractional times together that would be 22+46+72=140, then divide the 1st fraction by this sum to get: 22/140=.1571 or about 16%, then the 2nd fraction would be: 46/140=.3285 or about 33%, and finally: 72/140=.5142 or about 51%. So you would have 16%, 33% and 51% (=100% of the variant) rather than 17%, 33%, and 50% (100% of the variant, also) as was suggested in the earlier post. I would think that this might be a more accurate way of distributing the variant from track to track and distance to distance.

What do you think?

PS. If someone is using this method, I apologize for putting it "out there".

The only real way to make quality pace call variants is to do them separately. You will get some huge errors trying to use the final variant, or trying to break the final variant into fractions. It just doesn't work. I'm not trying to sound like a smart ass, but if you go down this road, you are going to waste a lot of time and have poor pace figures to show for it.

[raybo]

The only real way to make quality pace call variants is to do them separately. You will get some huge errors trying to use the final variant, or trying to break the final variant into fractions. It just doesn't work. I'm not trying to sound like a smart ass, but if you go down this road, you are going to waste a lot of time and have poor pace figures to show for it.

CJ, no problem. I value your advice and am sure you are correct. However, I'm not making pace figures. I'm only seeking to adjust fractional times from track to track. The variant discussion was a result of this goal. I realize that in order to adjust a horse's PP fractions, and be as accurate as possible, in relation to today's track, I need a complete set of fractional pars and, to be thorough in so doing, fractional variants should be applied to each PP before normalizing them to today's track. I won't be using them for calculations but rather for better visualization and comparison with others in the race.

I do have track pars with fractional times, but, I don't have the time or databasing skills to create fractional variants as you suggest.

[gm10]

I'd thought of that but am just in the initial stages of this line of thought. If variants are to be used for the fractional times as well as the final times, the question remains, how would you go about distributing the variant among the fractional times?

Would it be ok to use the distribution suggested earlier in the thread (17%, 33%, 50%)? This approach would seem to be too simplistic to me, surely every track and each distance would not have the same distribution for each fraction. On heavier tracks and at longer distances, logic would say, the track variant would affect the later fractions more than the early ones.

The following is "right off the top of my head" so bear with me.

How about finding the percentage of the "fraction vs the sum of the fractions" and then use the same percentage of the final time variant for that particular fraction?

ie; Let's assume that a race was run with the fractional times of 22 46 72. If you add the 3 fractional times together that would be 22+46+72=140, then divide the 1st fraction by this sum to get: 22/140=.1571 or about 16%, then the 2nd fraction would be: 46/140=.3285 or about 33%, and finally: 72/140=.5142 or about 51%. So you would have 16%, 33% and 51% (=100% of the variant) rather than 17%, 33%, and 50% (100% of the variant, also) as was suggested in the earlier post. I would think that this might be a more accurate way of distributing the variant from track to track and distance to distance.

What do you think?

PS. If someone is using this method, I apologize for putting it "out there".

It's an interesting idea. I've used 1/3 for each of the three fractions. This may be inaccurate but remember that I comparing this with performances who have also been DTV-adjusted this way. I guess this will neutralize quite a bit of DTV-fractional noise this way, though it also could end up flattening truly exceptional fractions. Any thoughts?

[gm10]

Are front runners winning a majority of the races? The above might indicate some type of speed bias.

mmm
very thought-provoking comment
could one quantify daily track bias by splitting up the DTV in fractional DTV's?

[classhandicapper]

I agree with everything CJ and Premier Turf Club had to say on this subject.

Obviously, wind, run up changes, crazy front runners, super slow paced races, the track not being uniformly fast in different areas etc... make it more difficult to calculate an accurate pace variant than final time variant.

I'm probably the most repetitive guy here (other than zilzal :lol: ), but this is why I advocate a multi-faceted approach.

If you watch races, look at the race development, look at how horses finished relative to what you expected, look at how the first ones back run etc.... that sometimes clarifies the complexities of numerical analysis and tells you when the numbers and formulas you are using may be a bit off. Sometimes even with extraordinary insights the figure maker will wind up with numbers that simply don't make much sense based on what you know about the horses.

The fat_man, (I believe currently barred from here) was a very strong advocate of the latter approach and produced some nice graphic illustrations of race development.

IMO, a blended approach produces the best results, but of course it's more work.

[Tom]

Class, I respect your approach, and we do what we believe in, but if you are not comprehensive, and go for volume, you will hit the ugly longshots that comprehension will steer you off.

When I have 3rd of a layoff, 4 rules horse with the best early rating, I don't look at trainer, jockey, weight, and I get really easy on class, and distance.

[Charlie D]

Unless, that is, the variant maker is aware of that issue and takes it into account.


But then it's no longer a track variant is it

[cj]

But then it's no longer a track variant is it

Of course it is. You are still measuring the speed of the track. The difference on other surfaces is that you have to find the segment(s) of the race where the real running took place.

[Charlie D]

How is it??

You are adding subjectivity to an already subjective measurement

[cj]

How is it??

You are adding subjectivity to an already subjective measurement

If it is already subjective, and I agree that it is, how does adding a little more subjectivity make it not a track variant? Is there some formula I don't know about to determine when the line is crossed?

You will never get an argument from me that making variants is an exact science. It is far from it, and tough work as well. That said, I have learned that adding expert subjectivity to the process makes the figures much better than sticking to a purely mechanical method. It doesn't mean I or any other figure maker gets them all right. That said, BRIS sticks with mechanical methods and their figures are pretty darn poor.

[Charlie D]

If it is already subjective, and I agree that it is, how does adding a little more subjectivity make it not a track variant? Is there some formula I don't know about to determine when the line is crossed?

.


:) fair comment CJ


Maybe it should be called a race variant

[classhandicapper]

Class, I respect your approach, and we do what we believe in, but if you are not comprehensive, and go for volume, you will hit the ugly longshots that comprehension will steer you off.

When I have 3rd of a layoff, 4 rules horse with the best early rating, I don't look at trainer, jockey, weight, and I get really easy on class, and distance.

Tom,

I agree with you.

If I think I've found some hidden value on a horse that doesn't have my top figures etc... I still tend to use it. I may just structure my bets differently.

I guess I'm talking more from a theoretical perspective about trying to get the best possible understanding of the performances of horses within a certain race. A dual approach can sometimes confuse you, sometimes clarify things etc... but sometimes confusion can be better than being wrong.

[delayjf]

The 17/33/50 number only applies to 6 furlongs; I would expect those numbers to change as the length of the final fraction increases. Those percentages do not apply to Routes, Turf, or Poly. You would have to compare the win pars to come up with the correct percentages and in cases of WX, no doubt the relationship can change.
How about finding the percentage of the "fraction vs the sum of the fractions" and then use the same percentage of the final time variant for that particular fraction?
I know Game Theory who posts here from time to time does exactly that and I know of at least one other well respected software developer who does the same thing. Given my respect for both, I’d say it would work for you. I know of some handicappers (including Brohammer, who uses them today) who now use Quirin pace figures which are (imho) less sophisticated than using Beyer Speed figures with variants made from pars alone.

I am not saying that either method above would be better than the more customized manner that CJ uses, if you can factor in wind, uneven track moisture content, changing gate placement – you’re the man. But given the success of other on this board and others I know of you could make it work.

[raybo]

I know Game Theory who posts here from time to time does exactly that and I know of at least one other well respected software developer who does the same thing. Given my respect for both, I’d say it would work for you. I know of some handicappers (including Brohammer, who uses them today) who now use Quirin pace figures which are (imho) less sophisticated than using Beyer Speed figures with variants made from pars alone.

I am not saying that either method above would be better than the more customized manner that CJ uses, if you can factor in wind, uneven track moisture content, changing gate placement – you’re the man. But given the success of other on this board and others I know of you could make it work.

Well, it's good to know that this approach might have some validity. I think I'll proceed along that line, for now anyway, and see what develops.

My overall goal, still evolving, is to adjust all the PPs for today's track, surface and distance.

Once I get those formulas in my program I hope to be able to scan each horse's PPs and spot both current form cycle placement and current performance ability, at today's distance, surface and track. This should enable me to better select proper running lines which, in turn, will allow me to plot each horse's selected running line on a chart so I can see the probable race shape.

Once I can accurately estimate the pace of the race I'll be in much better shape to determine top of stretch positioning and final run capabilities for each horse.

[Tom]

Here's what I do.
Take all the races at a given distance/surface/track condition. The more the merrier, but keep to the last couple of years, maybe three. Going back too far can hurt you if there have been changes. Make a scatter diagram using final time and second call time. Determine the trend line and the formula for it, ie, y=ax+b variety. Do the same with final time and first call time.

Make a "speed chart" based on final time. Column C is the final time, cloumn B is the results of y=ax+b for second call times associated with that final time, and column A is the same using first call time. Now, for every change in final time of a fifth ( or tenth) the chart tells me what to expect for a change in columns A and B.

As window dressing for this, I make a speed rating for each line on the chart, with final time as the reference point. This will tell me if a pace line was fast or slow at either pace call by looking at the speed ratings.

Now, what is significant as fast or slow?

I use Randy Giles' idea of a multiplier to let the data tell me that,too.

Using the same spreadsheet, calculate the percentage each second time is for the associated final time. Ie, 46 112 = .6388. Do the same for the first call times. Then, calculate the mean and standard deviation for each colum of percentages. Mean +/- 1 St Dev is normal, anything past 1 St Dev is fast or slow.

[rufus999]

I remember back in the '60's when Bob Hayes challenged a thoroughbred to a 40 yard dash. He lost. I don't believe any sort of trumped up fractional variant would have helped his sorry ass if he had chosen to try it again. I kinda think he felt the same way.:lol:

rufus:9::9::9:

[bobphilo]

I used to make pace variants for three calls, and a lot of times a lone front runner who ran exceptionally fast early on and then died threw the early variants out of wack.

Actually Tom, the opposite situation, where there is a slow pace, illustrates why it's dangerous making seperate pace and final time varaints. If the early fractions are slow, how does one know if it's because the track is slow or because the leader is being strongly rated. Early rating introduces an additional confouding variable.

Bob