Putting an end to Beyer type speed ratings

[traynor]

Originally Posted by TrifectaMike

What I am thinking of is a time measure based metric, which doesn't require distance or variant adjustment...etc.

Mike

Something like a parallel speed chart, in which X time at Y distance has a numeric value equivalent to M time at N distance so the values can be compared directly without any other adjustment?

[Maximillion]

I havent been here that long but from lurking and reading prior posts, I would
bet my entire bankroll that cjs figures are "better" than Beyers..so my question to cj would be....Is there a reason(s) for the change?

[Bullet Plane]

Originally Posted by Maximillion

I havent been here that long but from lurking and reading prior posts, I would
bet my entire bankroll that cjs figures are "better" than Beyers..so my question to cj would be....Is there a reason(s) for the change?

My best guess would be the incredible shrinking Beyers assosiated with Pars that might be suspect due to increased purses at Racino's.

[TrifectaMike]

Originally Posted by traynor

Something like a parallel speed chart, in which X time at Y distance has a numeric value equivalent to M time at N distance so the values can be compared directly without any other adjustment?

Attempt to apply this approach so that the effect of the surface speed is not observable nor necessary.

Mike (Dr Beav)

[Maximillion]

Originally Posted by Bullet Plane

My best guess would be the incredible shrinking Beyers assosiated with Pars that might be suspect due to increased purses at Racino's.

Guessing(I could be wrong) that cj does not use "pars"

[cj]

Originally Posted by Maximillion

I havent been here that long but from lurking and reading prior posts, I would
bet my entire bankroll that cjs figures are "better" than Beyers..so my question to cj would be....Is there a reason(s) for the change?

Yes...the figures were good, but I learned and found what I consider flaws in the Beyer methodology.

[cj]

Originally Posted by Maximillion

Guessing(I could be wrong) that cj does not use "pars"

No, I don't.

[cj]

Originally Posted by traynor

Without revealing any proprietary information, could you expand on your method of computing variants? That seems to be an area that needs serious revision in most approaches to speed and pace analysis, and I am curious how you resolved (are resolving) it.

Again, I am not asking you to give away the candy store, but more a brief explanation of your conceptual approach to making variants.

Well, the first step was fixing what I think are faulty speed charts, which really helps with the variant making. This is particularly true on different surfaces. I studied the layout and run up distance of every track in North America to know all the nuances of each distance and accounted for all of them.

Second, I incorporated the pace into an overall performance rating for all horses that were either competitive or a part of the pace. I already did this to a point, but now, I include more than just the winner. It gives me a much larger sample size for creating a variant. It also makes it easier to understand when very fast and very slow paces are the cause of slower than expected times.

While there are no guarantees, the variant making process has become MUCH easier and less time consuming. There aren't nearly the number of puzzling race cards and/or individual races as I've had in the past. It tells me I've made some very good changes.

[Maximillion]

Originally Posted by cj

Yes...the figures were good, but I learned and found what I consider flaws in the Beyer methodology.

Math isnt my strong point at all..(I try to avoid it)..but i would guess that the "beaten lengths" could be the "flaws" in the Beyer methodology?

[traynor]

Originally Posted by cj

Well, the first step was fixing what I think are faulty speed charts, which really helps with the variant making. This is particularly true on different surfaces. I studied the layout and run up distance of every track in North America to know all the nuances of each distance and accounted for all of them.

Second, I incorporated the pace into an overall performance rating for all horses that were either competitive or a part of the pace. I already did this to a point, but now, I include more than just the winner. It gives me a much larger sample size for creating a variant. It also makes it easier to understand when very fast and very slow paces are the cause of slower than expected times.

While there are no guarantees, the variant making process has become MUCH easier and less time consuming. There aren't nearly the number of puzzling race cards and/or individual races as I've had in the past. It tells me I've made some very good changes.

Something that few may have realized is that the Quirin approach to par times was intended to enable chart generation from a single value for each track. That is how all those neat "fifth of a second per class level" figures came about. That is, accumulations of data may be corrupted by that influence.

Making your own track charts is a chore, but likely well worth the effort.

Someone, somewhere (possibly Quirin, but I don't recall who it was offhand) suggested picking a mid-point between the winner and placer as the "real" final time of a race. I think considering multiple finishers in calculating time is a really sound idea, especially if adapted to varying gaps between the finishers (or main contenders for pace figures).

One of the best techniques I have found is limiting the slow but (generally) accepting the fast for daily variants. That is, there are few reasons for horses to run faster than normal, but a whole truckload of reasons (including being just naturally slower as a group) for running slower than "normal." It is an area that it might be worth your time to look at, because it can make a major difference in outcomes.

Another suggestion that may be worth a look is a technique developed (I think) by Jay Hovdey for harness programs in Canada. Rather than adjusting by fifths of a second or even hundredths of a second, he divided the actual by par to generate a value like 0.9978 or 1.0023 ( {"N4"} ), then used that to adjust incremental values. It eliminates the chore of dividing the variant to apply it to fractional times for pace calculations. It also works great for applying track-to-track adjustments, especially if you code in accumulators that let you make your own internal fraction pars.

[traynor]

Originally Posted by TrifectaMike

Attempt to apply this approach so that the effect of the surface speed is not observable nor necessary.

Mike (Dr Beav)

That sounds really interesting, although I don't have a clue how it would work. There was a technique the Sartin people used for awhile that they called a "meridian variant." In esssence, they averaged a bunch or races and came up with a "race profile." Every horse was calculated as having run its race in the "profile" times. There was a lot of marketing nonsense that went with it, but basically it was "use a static set of times for all entries, and only consider beaten lengths relative to that set of times to calculate ratings for individual entries."

The best thing that could be said about it is that it saved a lot of time entering data into their apps, because the entry that won or finished closest to the winner in the race used as a paceline always came out on top. (Except in UltraScan, which is another thing entirely.)

[MPRanger]

Originally Posted by Capper Al

What's wrong with Beyers other than they are bet down?

Speed figures based on par times adjusted for variants are excellent tools in
the right hands imho. Horses do not push the envelope very much. They
don't often set new records. Therefore, they generally run within the
parameters of the pars.

A horse does not run the same figure every time. This is the crux of it. There
is a reason for the figure he turns in. A horse with a lower figure in a
competitive early pace may turn in a slower figure than a lesser horse who
got loose on the lead. So just a higher number can be deceiving.

The same type variations in performance must occur in any numerical
approach to grading a horses perfomance.

Instead of a race being a contest of speed, I see it as a fight. A contest of
endurance. How a horse fights to stay on the pace will affect his final figure.
Lower speed figures with a finish in the top three or within 5 lengths is the
sign of a gutsy race horse in the claiming ranks. I wouldn't judge a drop
down horse in an allowance race this way though. So speed figures mean
nothing unless used in the proper context. But used correctly in each
scenario they can give you insight that the public looking at the same
numbers doesn't get.

To me, Beyer numbers are good enough even if something better never
comes along. But if it does, then I'll take a look at it. I would make sure I
was absolutely, positively using the available speed figures to their best
function before I invented something new.

[cj]

Originally Posted by Maximillion

Math isnt my strong point at all..(I try to avoid it)..but i would guess that the "beaten lengths" could be the "flaws" in the Beyer methodology?

I haven't used the Beyer beaten lengths formulas for a long time, so while I agree, that isn't a new change for me.

[highnote]

There are a lot of different ways to skin a cat. I'd recommend reading "Handicapping Speed" by Charles Carroll for some useful insight into speed figures. And Cary Fotias' book, "Blinkers Off" is also worth a read.

James Quinn's book, "Figure Handicapping" is excellent and he has a great way of making turf figures.

I'd recommend Len Ragozin's book, too. He doesn't tell you everything about how he makes his figures, but he gives some useful tips.

He says that Saratoga is a casual meet and that sometimes the starting gate is not always in the same position for a given distance. So he uses video to time the races and he knows the distance of the gate to the timer by looking at the starting gate in relation to the rail. (at least I think he said it was the rail?)

[dav4463]

Any good set of figures can help you narrow down true contenders. In most races though, there are quite a few horses capable of running fast enough to win. There are no figures that can tell you what a horse is going to do ....TODAY....No figure can tell you how a horse feels, or how the jockey feels, or can predict whether or not a horse gets a good start or gets boxed in, etc....that's what makes it fun!