Drosselmeyer BSF?

[gm10]

If I'm not mistaken, that was the second slowest Belmont Stakes in the last 40 years (the surface was quite slow, it has to be said).

I gave Drosselmeyer 76 (compared to 79 in the Dwyer).
That should convert into a two-digit BSF.

Weak crop, weak race. The older horses are more interesting this year.

[Jasonm921]

I don't do my own speed figures so I'm just guessing but I would be shocked to see anything more than an 87 or 88. This race was slow.

[depalma113]

If I'm not mistaken, that was the second slowest Belmont Stakes in the last 40 years (the surface was quite slow, it has to be said).

I gave Drosselmeyer 76 (compared to 79 in the Dwyer).
That should convert into a two-digit BSF.

Weak crop, weak race. The older horses are more interesting this year.


How about Pathetic.

[FenceBored]

94

http://www.drf.com/stakeresults/drfStakeResults.jsp

(http://www.drf.com/stakeresults/drfStakeResults.jsp)

[Jasonm921]

Outside of Bribon....looks like slow numbers all day. I thought Winchester would get at least a 105-107.

[Thomas Roulston]

I swept the 1 1/2-mile dirt stakes at Belmont this weekend (with Alcomo and Drosselmeyer).

[gm10]

94

http://www.drf.com/stakeresults/drfStakeResults.jsp

(http://www.drf.com/stakeresults/drfStakeResults.jsp)

cheers big ears

[cj]

A 94 is not pretty. I'm guessing First Dude is probably the best of those in this race over the long haul.

[dansan]

I'm sure andy and his clan will inflate the bsf just because it was the belmont

[cj]

I'm sure andy and his clan will inflate the bsf just because it was the belmont

You think a 94 is inflated?

[rastajenk]

I think he's predicting a change down the road. You know, like it happens all the time. :rolleyes:

[Greyfox]

Is testing for steroids is having an influence on the strength of the breed??

[dansan]

did they already give out bsf for the belmont?

[rastajenk]

See post #4.

[gm10]

You think a 94 is inflated?

I think it is a little bit tbh. If you look at all mile and a half races at Belmont on a fast track this century, this is the slowest race. 4 seconds slower than Summer Bird, five seconds slower than Point Given. Shockingly bad.

[westny]

Is testing for steroids is having an influence on the strength of the breed??

Very good point....added to the slow Belmont track..the horses were "buried into the track over their ankles"

The track was watered all day trying to combat the very deep track.

[cj]

I think it is a little bit tbh. If you look at all mile and a half races at Belmont on a fast track this century, this is the slowest race. 4 seconds slower than Summer Bird, five seconds slower than Point Given. Shockingly bad.

I haven't looked at it yet, but to be honest, the 1 1/2 figure for Belmont is pretty much impossible to make. The pace varies wildly and it is the only race run around a large section of the track. With the sheer size of Belmont the speed of different sections of the track can vary greatly.

[Steve R]

Completely wrong. This is another classic example of a continuously changing variant throughout the card. Using traditional fifth-of-a-second variants the pattern looks like this for the five dirt races:

Race 4, older horse NW1: time 1:22.0, par 1:22.3, variant 3 fast
Race 6, 3yo G2: time 1:22.3, par 1:22.0, variant 3 slow
Race 8, older horse G2: time 1:09.3, par 1:08.1, variant 7 slow
Race 9, 3yof G1: time 1:37.2, par 1:35.3, variant 9 slow
Race 11, 3yo G1: time 2:31.2, par 1:48.2, variant 15 slow

Plot the race number vs the race variant, apply linear regression analysis and the correlation coefficient is an exceptional 0.998, which confirms the continuous rate of change in track speed. You get essentially the same result if you plot the starting time of the race vs the variant.

If you apply the average variant, which is 6 slow, the Belmont time adjusts to 2:30.1. That is 9 ticks slower than the Belmont par. That is worth about 10 Beyer points at 12f. Beyer's par for the Belmont is 106. Roughly, then, the adjusted Beyer would be 96 - very close to his actual assignment of 94.

I have no idea how Beyer actually arrived at his number, but the concept of a variant changing from race to race simply eludes him (and almost all other figure makers). IMO this was an average Belmont if you consider all the dirt races or a bit substandard if you only consider the two races beyond traditional sprint distances. By my system, using the changing variant protocol, First Dude ran within a length of his Preakness performance and Stately Victor within a length of his Derby performance. The other two Derby starters, Ice Box and Make Music for Me, each regressed 10 lengths more or less off their Derby efforts. Drosselmeyer clearly moved forward as his pedigree was especially suited to the 12f, although I had his Louisiana Derby not too far below his Belmont.

[JustRalph]

wow......... bring a flame thrower into every room and eventually you are going to burn the house down around you.............

[cj]

Completely wrong. This is another classic example of a continuously changing variant throughout the card. Using traditional fifth-of-a-second variants the pattern looks like this for the five dirt races:

Race 4, older horse NW1: time 1:22.0, par 1:22.3, variant 3 fast
Race 6, 3yo G2: time 1:22.3, par 1:22.0, variant 3 slow
Race 8, older horse G2: time 1:09.3, par 1:08.1, variant 7 slow
Race 9, 3yof G1: time 1:37.2, par 1:35.3, variant 9 slow
Race 11, 3yo G1: time 2:31.2, par 1:48.2, variant 15 slow

Plot the race number vs the race variant, apply linear regression analysis and the correlation coefficient is an exceptional 0.998, which confirms the continuous rate of change in track speed. You get essentially the same result if you plot the starting time of the race vs the variant.

If you apply the average variant, which is 6 slow, the Belmont time adjusts to 2:30.1. That is 9 ticks slower than the Belmont par. That is worth about 10 Beyer points at 12f. Beyer's par for the Belmont is 106. Roughly, then, the adjusted Beyer would be 96 - very close to his actual assignment of 94.

I have no idea how Beyer actually arrived at his number, but the concept of a variant changing from race to race simply eludes him (and almost all other figure makers). IMO this was an average Belmont if you consider all the dirt races or a bit substandard if you only consider the two races beyond traditional sprint distances. By my system, using the changing variant protocol, First Dude ran within a length of his Preakness performance and Stately Victor within a length of his Derby performance. The other two Derby starters, Ice Box and Make Music for Me, each regressed 10 lengths more or less off their Derby efforts. Drosselmeyer clearly moved forward as his pedigree was especially suited to the 12f, although I had his Louisiana Derby not too far below his Belmont.

I do the same thing in many cases, but there is no way to know for sure a track the size of Belmont is uniform speed wise all the way around.

[Tom]

So yo have a BSF for the Belmont...whatcha gonna do with it?
Me, I will never use it or that pace numbers for anything.
Unless they run another 12 furlongs someday.

The word speed and this races seem to not go together.

[cj]

So yo have a BSF for the Belmont...whatcha gonna do with it?
Me, I will never use it or that pace numbers for anything.
Unless they run another 12 furlongs someday.

The word speed and this races seem to not go together.

I agree, it is really not very valuable at all for the future.

[gm10]

Completely wrong. This is another classic example of a continuously changing variant throughout the card. Using traditional fifth-of-a-second variants the pattern looks like this for the five dirt races:

Race 4, older horse NW1: time 1:22.0, par 1:22.3, variant 3 fast
Race 6, 3yo G2: time 1:22.3, par 1:22.0, variant 3 slow
Race 8, older horse G2: time 1:09.3, par 1:08.1, variant 7 slow
Race 9, 3yof G1: time 1:37.2, par 1:35.3, variant 9 slow
Race 11, 3yo G1: time 2:31.2, par 1:48.2, variant 15 slow

Plot the race number vs the race variant, apply linear regression analysis and the correlation coefficient is an exceptional 0.998, which confirms the continuous rate of change in track speed. You get essentially the same result if you plot the starting time of the race vs the variant.

If you apply the average variant, which is 6 slow, the Belmont time adjusts to 2:30.1. That is 9 ticks slower than the Belmont par. That is worth about 10 Beyer points at 12f. Beyer's par for the Belmont is 106. Roughly, then, the adjusted Beyer would be 96 - very close to his actual assignment of 94.

I have no idea how Beyer actually arrived at his number, but the concept of a variant changing from race to race simply eludes him (and almost all other figure makers). IMO this was an average Belmont if you consider all the dirt races or a bit substandard if you only consider the two races beyond traditional sprint distances. By my system, using the changing variant protocol, First Dude ran within a length of his Preakness performance and Stately Victor within a length of his Derby performance. The other two Derby starters, Ice Box and Make Music for Me, each regressed 10 lengths more or less off their Derby efforts. Drosselmeyer clearly moved forward as his pedigree was especially suited to the 12f, although I had his Louisiana Derby not too far below his Belmont.

I do the same as you do but I don't really see the same. It's more like a V-curve, with races 1-2 (and nearly 4) close to par, then becoming quicker, but slowing down for the last two races again. Same when I use the Equibase par times. Not sure if there's any reason for such a DTV-curve. Maybe they did some track maintenance in between races.

[Tom]

Take the margins into account and the F3 race looks a lot slower. Winner by 4, place horse by 7 ....no way I give that a F3. And using the Belmont for your variant is not a good idea IMHO. That leaves a S3,S7,S9.

[ezpace]

They were putting 10,000 gals of water on the
track in between races .Hard to get it uniform when it's drying out
so quickly. source ESPN coverage
Espn did a super helpful job of track condition info throughout
the day.

[Tom]

They upped it to 12,000 prior to the Belmont.

[Steve R]

I do the same as you do but I don't really see the same. It's more like a V-curve, with races 1-2 (and nearly 4) close to par, then becoming quicker, but slowing down for the last two races again. Same when I use the Equibase par times. Not sure if there's any reason for such a DTV-curve. Maybe they did some track maintenance in between races.
Even if you use something as crude as the average times of the last five or ten runnings of the Stephens, True North, Acorn and Belmont as approximations of par times, the correlation coefficient is still around 0.9. Track maintenance is surely an important factor but no more important than changes in surface temperature, wind or humidity, and these can change constantly throughout the day. In any case, it is quite clear that applying a uniform variant under Saturday's conditions is a gross error. Of course it's easy to justify an awful performance if you go in with the idea that the horses suck. But when I consider First Dude's Preakness and Drosselmeyer's Louisiana Derby as performance benchmarks combined with each horse's suitability to the distance, I don't see how the deteriorating track speed over the day can be dismissed. Physics trumps projection every time.

[gm10]

Even if you use something as crude as the average times of the last five or ten runnings of the Stephens, True North, Acorn and Belmont as approximations of par times, the correlation coefficient is still around 0.9. Track maintenance is surely an important factor but no more important than changes in surface temperature, wind or humidity, and these can change constantly throughout the day. In any case, it is quite clear that applying a uniform variant under Saturday's conditions is a gross error. Of course it's easy to justify an awful performance if you go in with the idea that the horses suck. But when I consider First Dude's Preakness and Drosselmeyer's Louisiana Derby as performance benchmarks combined with each horse's suitability to the distance, I don't see how the deteriorating track speed over the day can be dismissed. Physics trumps projection every time.

Sure, but you aren't using physics. You see a numerical trend and you assume that physics are behind that trend. I'm not really going to argue for or against a uniform DTV. I use a local polynomial regression for my DTV - so it is certainly not uniform and it would bring out a linear trend if there was one, but I'm not really seeing the consistently slower times. Maybe your or my par times are a bit off. Who knows - it was certainly not very good.

[Steve R]

Sure, but you aren't using physics. You see a numerical trend and you assume that physics are behind that trend. I'm not really going to argue for or against a uniform DTV. I use a local polynomial regression for my DTV - so it is certainly not uniform and it would bring out a linear trend if there was one, but I'm not really seeing the consistently slower times. Maybe your or my par times are a bit off. Who knows - it was certainly not very good.
I never said I was using physics to measure the track speed. What I was trying to convey was the notion that the actual physical changes in the racing surface preempt any attempt to project what the speed of the race should be based on one's perceptions.

Could you post your par times for the four dirt stakes races and the allowance? I'd like to see how far apart we are. Also, when I run a second-order polynomial regression on the data, I get the same result. Either way, the correlation coefficients are essentially 1.00.

Maybe I'm missing something, but I thought r-squared values above 0.99 in the analyses presented earlier meant that 99% of the variation in individual race variant is attributable to the time of day the race was run. If the r-squared values were generally low I would agree that such a conclusion was questionable. But that's not the case here.

[gm10]

I never said I was using physics to measure the track speed. What I was trying to convey was the notion that the actual physical changes in the racing surface preempt any attempt to project what the speed of the race should be based on one's perceptions.

Could you post your par times for the four dirt stakes races and the allowance? I'd like to see how far apart we are. Also, when I run a second-order polynomial regression on the data, I get the same result. Either way, the correlation coefficients are essentially 1.00.

Maybe I'm missing something, but I thought r-squared values above 0.99 in the analyses presented earlier meant that 99% of the variation in individual race variant is attributable to the time of day the race was run. If the r-squared values were generally low I would agree that such a conclusion was questionable. But that's not the case here.

R2 tells you how much of your variance is being explained by your model. In this case, you would be be predicting the discrepancy between par times and actual times on the basis of race number or post time. From your reply I take it that you are using post times. R2 of 0.99 is exceptional in real linear regression analysis, I have to say. Can you tell me the intercept and slope of your analysis - and what your X and Y are?

Simple linear regression use a simple line/plane/hypercube as a model, polynomial regression is a non-parametric model. In the case of DTV modeling, I prefer polynomial fitting, as the Gaussian conditions which linear regression relies on, are often not satisfied.

Here are the sets of par times ...

RACE_NUMBER PARTIME PARTIME_EQUIBASE
1 85.389 85.51
2 72.489 72.66
4 83.859 83.15
6 83.139 82.36
8 69.969 69.32
9 96.837 95.67
11 151.184 N/A

[Jasonm921]

I think they made three trips with the water truck between the Manhattan and the Belmont Stakes. They tried to saturate it.

[Tom]

Your sample size to too small. You have to make the assumption that every race ran the same from par, which may or may not be true. The wire to wire 9th may have been slow in the middle part, as I have it fast slightly for 4 furlong, slow a second at the 6 furlongs, so that final slow time may not have been as bad as it looked.

GM, why two par times?

[Steve R]

Take the margins into account and the F3 race looks a lot slower. Winner by 4, place horse by 7 ....no way I give that a F3. And using the Belmont for your variant is not a good idea IMHO. That leaves a S3,S7,S9.
Even if the allowance is as much as S1 the correlation coefficient drops only from 0.996 to 0.980. Leave it out entirely and it's 0.993. Also, I don't understand your issue with the Belmont. The average time over the last ten runnings on a fast track is 2:28.4. Since 1983 it's 2:28.2. IOW, it hasn't changed much in over 25 years, so I would consider that a reasonable par time from which to calculate a variant. Leave out both the allowance and the Belmont and the correlation coefficient improves to 1.000. No matter how you consider it, the track was simply getting slower all day.

[Steve R]

R2 tells you how much of your variance is being explained by your model. In this case, you would be be predicting the discrepancy between par times and actual times on the basis of race number or post time. From your reply I take it that you are using post times. R2 of 0.99 is exceptional in real linear regression analysis, I have to say. Can you tell me the intercept and slope of your analysis - and what your X and Y are?

Simple linear regression use a simple line/plane/hypercube as a model, polynomial regression is a non-parametric model. In the case of DTV modeling, I prefer polynomial fitting, as the Gaussian conditions which linear regression relies on, are often not satisfied.

Here are the sets of par times ...

RACE_NUMBER PARTIME PARTIME_EQUIBASE
1 85.389 85.51
2 72.489 72.66
4 83.859 83.15
6 83.139 82.36
8 69.969 69.32
9 96.837 95.67
11 151.184 N/A
The results for second-order polynomial regression and linear regression are virtually identical with r-squared >0.99 in both cases. The same is true for race number or post time vs variant. For race number vs variant the linear equation is Variant = 18.93*Race# + 90.48. In the polynomial regression case the equation is Variant = 0.071*Race#^2 + 17.88*Race# - 86.97.

I don't know how Equibase calculates their pars and I use the pars from Cynthia Publishing modified by my own updates. In any case they have to be approximations in graded races because so few are run in a season. For example, prior to Saturday there had been only six Belmont graded races at 7f since the beginning of 2009 and only one for three-year-old males. There had been only three at 6f with two for older males and there had been 13 at 8f with just one for three-year-old fillies. And there were three 12f races with one for three-year-olds. IOW, the par times for these events are notional at best. FWIW, these are the average running times of the four graded races in question over the previous five years:

Woody Stephens, 1:21.1 (in 2009, 1:20.3)
True North, 1:08.2 (in 2009, 1:07.4)
Acorn, 1:35.0 (in 2009, 1:34.3)
Belmont, 2:28.0 (in 2009, 2:27.2)

I'd say these are as good as any to use as par times considering the paucity of these races at Belmont Park. If I do, and I include the Cynthia Publishing par time for the older horse allowance at 7f, I still get a correlation coefficient of 0.90. All of the Equibase par times are quite a bit slower than those of Cynthia Publishing and of the average times over the last few years while the last two are in reasonable agreement.

I use a different scale from Beyer and I incorporate pace into my single figure, but I have generated approximate equivalents to BSFs which correlate fairly well. So from a different angle, my changing variant analysis affords the following:

D' Funnybone, a 106 with an earlier 103.
Bribon, a 110 with an earlier 109.
Champagne d'Oro, a 101 with an earlier 97.
Drosselmeyer, a 103 with an earlier 102.

Each of these winning performances is consistent with previous top efforts in graded races. Of course everyone does it differently and I would love to hear Beyer's reasoning on this one.

[bisket]

A 94 is not pretty. I'm guessing First Dude is probably the best of those in this race over the long haul.
better than kensie last year? :lol:

[Cratos]

If I'm not mistaken, that was the second slowest Belmont Stakes in the last 40 years (the surface was quite slow, it has to be said).

I gave Drosselmeyer 76 (compared to 79 in the Dwyer).
That should convert into a two-digit BSF.

Weak crop, weak race. The older horses are more interesting this year.

You are correct the Belmont surface was very slow and tiring. Checklist the favorite in the 6f True North had a clear lead at the eighth pole and faded to finish 4th in a 1:09 and change final time.

My SSRV (surface speed resistance variant) is between a+1 to a+2 seconds slow without the wind being taken into consideration

[The Hawk]

CJ: Is there any way to make a reliable figure for the Belmont (any year) besides projecting the figure, based on the horses' past figures? At 1 1/2 miles, and with few routes on the main track (and even those are not run in close proximity to the Belmont), it seems crazy to do anything but make a stand-alone variant.

[bisket]

this is turning out to maybe be a good year for me. eventhough i missed a score yesterday; another day of questionable figs has fallen into my lap. on a big day of racing again to even add to a windfall later this summer and fall. i thought it would be a tough one this year now that the cup is off the plastic, but things are looking up :ThmbUp: just gotta get a new pair of lucky underwear.

[cj]

CJ: Is there any way to make a reliable figure for the Belmont (any year) besides projecting the figure, based on the horses' past figures? At 1 1/2 miles, and with few routes on the main track (and even those are not run in close proximity to the Belmont), it seems crazy to do anything but make a stand-alone variant.

I agree. There have, at times, been other 1 1/2 mile races run on Belmont day, but without them I don't see how. As I said, and was ignored, a track as big is Belmont is most likely not going to have a uniform surface all the way around. How anyone thinks they can project the variant from other races when around 5f are run on a part of the track not touched by a horse all day outside the Belmont is beyond me.

[cj]

this is turning out to maybe be a good year for me. eventhough i missed a score yesterday; another day of questionable figs has fallen into my lap. on a big day of racing again to even add to a windfall later this summer and fall. i thought it would be a tough one this year now that the cup is off the plastic, but things are looking up :ThmbUp: just gotta get a new pair of lucky underwear.

So loving a horse well in advance of the race and not making a dime (losing every bet actually) when he wins is turning this into a good year? One can only imagine how bad it has been up until now!

[Greyfox]

As I said, and was ignored, a track as big is Belmont is most likely not going to have a uniform surface all the way around.

I agree with you to a point.
Belmont is huge. The race was long.
Your comment begs the question,
"Why should any track, a mile or so, have a uniform surface all the way around?"

My suspicion is, they don't have a uniform surface , every day at least.
(Track maintenance is something that simulcast viewers never see. Machinations of the plow boys is something that should never happen on a given day. Yet....it does.)

[Tom]

When they use the whole Belmont tack, don't they go through New Jersey for part of the race?

[Show Me the Wire]

...............................Machinations of the plow boys is something that should never happen on a given day. [/i]Yet....it does.)


Especially in the midst of a mutli-race wager. I used to hate it when I saw the water trucks reverse direction ;)

But back to the topic. The Belmont was contested by a bunch of slow horses regardless of water trucks and/or tractors. I don't think paving would have helped this group.

[Steve R]

I agree. There have, at times, been other 1 1/2 mile races run on Belmont day, but without them I don't see how. As I said, and was ignored, a track as big is Belmont is most likely not going to have a uniform surface all the way around. How anyone thinks they can project the variant from other races when around 5f are run on a part of the track not touched by a horse all day outside the Belmont is beyond me.
The moving average of the Belmont final time has hardly changed in 20 years. How can the average Belmont time not be an acceptable par time from which to measure the variant? The Belmont varies around essentially the same time every year, some a few ticks faster, some a few ticks slower. If the average changed dramatically that would be a different story, but it doesn't. And why does it matter whether or not horses run over the same part of the track? If the par times (and fractional par times) are accurate, they incorporate all the differences in track surface associated with the various distances. Now if your point is that different parts of the track behave differently on any given day, then most times you couldn't calculate a variant at all. For example, on a card with two 5f sprints, two 6f sprints and two 7f sprints, the winners of the 5f races never saw the 6f and 7f parts of the track and the 6f winners never saw the 7f part of the track. So I guess you're left with the option of separate variants at each distance based on just two races each. That really makes the whole point of variants useless. But if you can measure identifiable patterns within the variants during the day, you are ahead of the game. Sometimes the pattern is no pattern at all, which suggests that an overall average will suffice. My experience is that variants change during the day more often than they don't, which is why I run regression analysis on every card. I think it's a much more credible approach than Beyer-type projection where, if you get one or two wrong, those that follow will be wrong as well.

[cj]

I agree with you to a point.
Belmont is huge. The race was long.
Your comment begs the question,
"Why should any track, a mile or so, have a uniform surface all the way around?"

My suspicion is, they don't have a uniform surface , every day at least.
(Track maintenance is something that simulcast viewers never see. Machinations of the plow boys is something that should never happen on a given day. Yet....it does.)

I agree, most don't most days. But they race over the whole strip at least a few times a day and the various sections are timed. That way you have something to go by.

[cj]

The moving average of the Belmont final time has hardly changed in 20 years. How can the average Belmont time not be an acceptable par time from which to measure the variant? The Belmont varies around essentially the same time every year, some a few ticks faster, some a few ticks slower. If the average changed dramatically that would be a different story, but it doesn't. And why does it matter whether or not horses run over the same part of the track? If the par times (and fractional par times) are accurate, they incorporate all the differences in track surface associated with the various distances. Now if your point is that different parts of the track behave differently on any given day, then most times you couldn't calculate a variant at all. For example, on a card with two 5f sprints, two 6f sprints and two 7f sprints, the winners of the 5f races never saw the 6f and 7f parts of the track and the 6f winners never saw the 7f part of the track. So I guess you're left with the option of separate variants at each distance based on just two races each. That really makes the whole point of variants useless. But if you can measure identifiable patterns within the variants during the day, you are ahead of the game. Sometimes the pattern is no pattern at all, which suggests that an overall average will suffice. My experience is that variants change during the day more often than they don't, which is why I run regression analysis on every card. I think it's a much more credible approach than Beyer-type projection where, if you get one or two wrong, those that follow will be wrong as well.

I am very leery of using pars in isolation. I like to combine that approach along with some projections. How can you compare some of the Belmont fields with others. Just on Beyer alone there have been some with Beyers approaching 120, and others with Beyers not cracking 100.

I understand about the 5-6-7f dilemma, but at least in that case they are running into the same direction of wind and on a part getting probably the same amount of sun.

[FEARTHECHOMP]

94 IS WEAK.....Even Da tara was better and he hasnt won a race since....

[gm10]

Your sample size to too small. You have to make the assumption that every race ran the same from par, which may or may not be true. The wire to wire 9th may have been slow in the middle part, as I have it fast slightly for 4 furlong, slow a second at the 6 furlongs, so that final slow time may not have been as bad as it looked.

GM, why two par times?

The first par times are my own (and my main source), the second set is for backup and verification.

[classhandicapper]

Now if your point is that different parts of the track behave differently on any given day, then most times you couldn't calculate a variant at all. For example, on a card with two 5f sprints, two 6f sprints and two 7f sprints, the winners of the 5f races never saw the 6f and 7f parts of the track and the 6f winners never saw the 7f part of the track. So I guess you're left with the option of separate variants at each distance based on just two races each. That really makes the whole point of variants useless.

EXACTLY!

Now throw in the issue of changing wind direction and intensity impacting various distances differently, run up changes, the track being dried out by the sun in one area but shaded in another, track maintenance between races, the impact of pace on time etc... and you begin to realize how debates (or even bets) over a couple of fifths of a second here are there are mostly a waste of time.

There are several very good figure makers out there. If you look at their figures for any given race you will see a few wide discrepancies and some smaller ones all over the place because of differences in methodology and interpretation of results.

Figures have an important role, but they can do more to steer you away from reality than enlighten you if you allow them to dictate your thinking about individual performances or a horse's ability.

[gm10]

The results for second-order polynomial regression and linear regression are virtually identical with r-squared >0.99 in both cases. The same is true for race number or post time vs variant. For race number vs variant the linear equation is Variant = 18.93*Race# + 90.48. In the polynomial regression case the equation is Variant = 0.071*Race#^2 + 17.88*Race# - 86.97.

I don't know how Equibase calculates their pars and I use the pars from Cynthia Publishing modified by my own updates. In any case they have to be approximations in graded races because so few are run in a season. For example, prior to Saturday there had been only six Belmont graded races at 7f since the beginning of 2009 and only one for three-year-old males. There had been only three at 6f with two for older males and there had been 13 at 8f with just one for three-year-old fillies. And there were three 12f races with one for three-year-olds. IOW, the par times for these events are notional at best. FWIW, these are the average running times of the four graded races in question over the previous five years:

Woody Stephens, 1:21.1 (in 2009, 1:20.3)
True North, 1:08.2 (in 2009, 1:07.4)
Acorn, 1:35.0 (in 2009, 1:34.3)
Belmont, 2:28.0 (in 2009, 2:27.2)

I'd say these are as good as any to use as par times considering the paucity of these races at Belmont Park. If I do, and I include the Cynthia Publishing par time for the older horse allowance at 7f, I still get a correlation coefficient of 0.90. All of the Equibase par times are quite a bit slower than those of Cynthia Publishing and of the average times over the last few years while the last two are in reasonable agreement.

I use a different scale from Beyer and I incorporate pace into my single figure, but I have generated approximate equivalents to BSFs which correlate fairly well. So from a different angle, my changing variant analysis affords the following:

D' Funnybone, a 106 with an earlier 103.
Bribon, a 110 with an earlier 109.
Champagne d'Oro, a 101 with an earlier 97.
Drosselmeyer, a 103 with an earlier 102.

Each of these winning performances is consistent with previous top efforts in graded races. Of course everyone does it differently and I would love to hear Beyer's reasoning on this one.

What are the actual Y-values? Just wondering about what your scale your variants are on.

Btw, I meant local polynomial regression which is a non-paramtetrical regression. Linear/quadratic regression tries to fit the data along a line/curve - local polynomial assumes no shape at all and lets the data govern its own shape. The conditions that need to be satisfied for it to work are much less stiff than for classical regression.

[PhantomOnTour]

What are the actual Y-values? Just wondering about what your scale your variants are on.

Btw, I meant local polynomial regression which is a non-paramtetrical regression. Linear/quadratic regression tries to fit the data along a line/curve - local polynomial assumes no shape at all and lets the data govern its own shape. The conditions that need to be satisfied for it to work are much less stiff than for classical regression.
Good Lord ya'll, we're just trying to get an accurate variant for the day. Put away all the algorythms and logorhythms and little bagariddims. Thought this was a thread about cleaning up the oil spill, what with all this linear regression and polynomial talk....uhhhhhh, I'm a liberal arts guy...sorry, carry on.

[Gorgeous George]

Good Lord ya'll, we're just trying to get an accurate variant for the day. Put away all the algorythms and logorhythms and little bagariddims. Thought this was a thread about cleaning up the oil spill, what with all this linear regression and polynomial talk....uhhhhhh, I'm a liberal arts guy...sorry, carry on.

Spelt 'algorithms' and 'logarithms'. You twaaaaaaaaaaaaaa.

[PhantomOnTour]

Spelt 'algorithms' and 'logarithms'. You twaaaaaaaaaaaaaa.
Both spellings appear on dictionary.com with a 'Y' or an 'I'. Like the word 'spelt', which seems odd but is actually a word. Almost Georgie, almost. You must still be upset over the Henry handball :lol: :lol: :lol:
Twaaaaaaaaaa

[Gorgeous George]

Both spellings appear on dictionary.com with a 'Y' or an 'I'. Like the word 'spelt', which seems odd but is actually a word. Almost Georgie, almost. You must still be upset over the Henry handball :lol: :lol: :lol:
Twaaaaaaaaaa

Does Logarithms have two 'o's in it? i didnt think so, and we have a winner, get in :jump:

You twaaaaaaaaaaaaaaaaa:D