Carroll: Handicapping Speed

[cj]

I've read this book and like a lot of what he writes. I know others here do as well. However, the baseline he uses is based on "World Records". I think it kind of falls apart there. He considers the following the best a horse can run, and thus equal on the speed scale:

Distance in furlongs, Time
4.0, 44.2
4.5, 50.4
5.0, 55.2
5.5, 61.4
6.0, 67.2
6.5, 73.6
7.0, 79.4
7.5, 86.6
8.0, 92.2
8.5, 98.4
9.0, 105.0
9.5, 112.4
10.0, 117.8

Why do I say it falls apart? Look at the difference adding an extra half furlong. For example, the first jump is 44.2 to 50.4, or 6.2. Here is the series:

6.2, 4.8, 6.2, 5.8, 6.4, 5.8, 7.2, 5.6, 6.2, 6.6, 7.4, 5.4

There is no way this can be an accurate baseline in my opinion. Horses would go from 9 to 9.5 furlongs in 7.4 seconds, then only take another 5.4 to run another 1/16 to make 10? Most of it doesn't make sense.

Any thoughts?

[Greyfox]

Sorry I haven't read the book. But just looking at the four furlong time I've seen lots of horses at Santa Anita run faster than 44.2 seconds.
I seem to be missing something in what he's saying.

[Robert Goren]

The problem is that some these distances are not run very often and they are run even less by good horses. For instance, the 7 1/2 Furlong distance is almost never run. 4 furlongs is never run by good older horses.

[cj]

Those are smaller problems, because the same inconsistencies exist among the commonly run distances.

[toetoe]

I love this guy and I'm pleasantly surprised that you like him, cj. He moved away from Las Vegas; not sure where he landed.

I say just disregard 9/16, 15/16 and 19/16 for handicapping purposes. That would not eliminate too many races, right ?

[maddog42]

Cj, you are absolutely right in your assessment of the World Record baseline.
But how can this method work at all? And it does work fairly well.

I came across a used copy of Carrolls Speed Handicapper V5 a few years ago.
At some tracks and distances it was Dead on. At others not so good. It was much better on dirt than turf and better on Sprints than routes. The percentage of World Record concept was not bad, so I tried using Percentage of Par. This was a slight improvement. One conclusion I came away with from the book and his program, was that MOST people over adjust their running lines. His program has a place for track to track adjustments and a place for daily variant adjustments.
IMO his method flatly does not work well above 8.5 or 9f. Throw out 4.5 furlong and lower and disregard 9f and above and you can see how things are greatly improved. It also works better on quarter horses than t-breds. I started making a notebook of class pars for every track in North America and was using his method and a 10k par. Thats right. Lotsa work. I was also using a calculator. My computer skills are lacking, or I would have pursued it farther.

[The Judge]

never knew he was in Vegas thought he stayed pretty much in New Mexico. I knew he like the quarter horses a lot.

http://desertsea.com/

[maddog42]

I've read this book and like a lot of what he writes. I know others here do as well. However, the baseline he uses is based on "World Records". I think it kind of falls apart there. He considers the following the best a horse can run, and thus equal on the speed scale:

Distance in furlongs, Time
4.0, 44.2
4.5, 50.4
5.0, 55.2
5.5, 61.4
6.0, 67.2
6.5, 73.6
7.0, 79.4
7.5, 86.6
8.0, 92.2
8.5, 98.4
9.0, 105.0
9.5, 112.4
10.0, 117.8

Why do I say it falls apart? Look at the difference adding an extra half furlong. For example, the first jump is 44.2 to 50.4, or 5.8. Here is the series:

5.8, 4.8, 6.2, 5.8, 6.4, 5.8, 7.2, 5.6, 6.2, 6.6, 7.4, 5.4

There is no way this can be an accurate baseline in my opinion. Horses would go from 9 to 9.5 furlongs in 7.4 seconds, then only take another 5.4 to run another 1/16 to make 10? Most of it doesn't make sense.

Any thoughts?

This baseline is based on FACTS. I am very critical of it also. A better question would be: Why do the World Records have these weird differences in
half furlong gaps ?

[cj]

This baseline is based on FACTS. I am very critical of it also. A better question would be: Why do the World Records have these weird differences in
half furlong gaps ?

It is probably because they are run on different tracks and surfaces with varying run ups.

[cj]

IMO his method flatly does not work well above 8.5 or 9f. Throw out 4.5 furlong and lower and disregard 9f and above and you can see how things are greatly improved.

Even in that narrow range it looks fishy:

5.5-6.0: 5.8
6.0-6.5: 6.4
6.5-7.0: 5.8

[maddog42]

Even in that narrow range it looks fishy:

5.5-6.0: 5.8
6.0-6.5: 6.4
6.5-7.0: 5.8

You have me playing the Devils Advocate here. I have problems with this baseline too. My win percentage with this method was 39-40 % (top 2)with a low mutuel. I was unable to get many overlays. My assertion was that it works better than you would think.

[maddog42]

It is probably because they are run on different tracks and surfaces with varying run ups.
Does that explains the differences? Maybe. I believe the deceleration and turns
might be a big factor. Runup? Certainly. This might be the biggest factor. Aren't you surprised that the World Records vary this much?

[proximity]

Sorry I haven't read the book. But just looking at the four furlong time I've seen lots of horses at Santa Anita run faster than 44.2 seconds.
I seem to be missing something in what he's saying.

the 4f final time baseline would potentially be slower (than some 4f fractions) because the race would generally start closer to a turn and go completely around a turn.

one thing that the book did make me think about was that the closer a horse gets to the true baseline, possibly (?) the larger the gap in 1/5 sec or 1 length would be. for example the difference between 6f in 106:3 and 6f in 106:4 would be greater (and even greater than beyer/ the sheets have it) than the difference between 6f in 113 and 6f in 113:1. :confused:

[pondman]

Why do I say it falls apart? Look at the difference adding an extra half furlong. For example, the first jump is 44.2 to 50.4, or 6.2. Here is the series:




You might want to average the furlongs into seconds:

Furlong, Seconds
4.0: 11.05
4.5: 11.20
5.0: 11.04
5.5: 11.16
6.0: 11.20
6.5: 11.32
7.0: 11.34
7.5: 11.54
8.0: 11.52
8.5: 11.57
9.0: 11.66
9.5: 11.83
10: 11.78

[cj]

You might want to average the furlongs into seconds:

Furlong, Seconds
4.0: 11.05
4.5: 11.20
5.0: 11.04
5.5: 11.16
6.0: 11.20
6.5: 11.32
7.0: 11.34
7.5: 11.54
8.0: 11.52
8.5: 11.57
9.0: 11.66
9.5: 11.83
10: 11.78

What does this show? There are still the same discrepancies.

[pondman]

What does this show? There are still the same discrepancies.

Easier to see a linear relationship.

These are base on stars-- horses such as housebuster and Best Pal. I don't like the methodology. How's he apply it to the daily claimer?

[cj]

Easier to see a linear relationship.

These are base on stars-- horses such as housebuster and Best Pal. I don't like the methodology. How's he apply it to the daily claimer?

I get that, but the relationship isn't really linear.

I think it would work fine for claimers. You have to establish a baseline somewhere, be it the high end, the middle, or the bottom. I've done all and they don't change things very much.

[TrifectaMike]

I get that, but the relationship isn't really linear.

I think it would work fine for claimers. You have to establish a baseline somewhere, be it the high end, the middle, or the bottom. I've done all and they don't change things very much.

Have you tried all three points as a baseline and fit straights lines between the points for the various distances and let the chart self scale?

Mike (Dr Beav)

[maddog42]

Have you tried all three points as a baseline and fit straights lines between the points for the various distances and let the chart self scale?

Mike (Dr Beav)

Are you talking about what Carroll did by plotting his World Records on a graph?
Carroll talks about the aberrations when he graphs his WR's. No one has mentioned on this thread about how Carroll fudges on his data. To be fair to the man he does cop to it in his book. When a very fast horse like Zany Tactics
runs on a very fast track like Turf Paradise, you get a WR of 1:06.4. Carroll disregards this "new" World Record and uses the old one, because it fits his straight line graph of WR times.

[cj]

Are you talking about what Carroll did by plotting his World Records on a graph?
Carroll talks about the aberrations when he graphs his WR's. No one has mentioned on this thread about how Carroll fudges on his data. To be fair to the man he does cop to it in his book. When a very fast horse like Zany Tactics
runs on a very fast track like Turf Paradise, you get a WR of 1:06.4. Carroll disregards this "new" World Record and uses the old one, because it fits his straight line graph of WR times.

He does indeed. What he doesn't tell is what records he used, and on which tracks they were run. He also doesn't mention that adding a second turn to the routes should skew the line a little, but it doesn't. There was definitely some fudgery and backfitting going on.

[Horseplayersbet.com]

Quick thoughts, 4 furlong races have a higher percentage of turn than any other sprint race, so it figures to have a slower furlong for second than 5 or 5 and a half does.
As for 1 mile and an eighth, if the track record is based on a one turn race I can see why there is a discrepancy.
Is he basing his times on a one mile track, or is he not taking this into consideration?

[The Judge]

I mentioned this horse not to long ago in the thread on Great Horses we have seen Simply Majestic wouldn't be considered a great race horse I don't think, but he was great on one day.

It was 1988 at Golden Gate Fields ran a 1:45 flat amazing. Jockey was Russell Baze.

Check this out and see if it adds to the conversation.

http://www.chef-de-race.com/articles/speed_in_the_thoroughbred.htm

[Native Texan III]

I've read this book and like a lot of what he writes. I know others here do as well. However, the baseline he uses is based on "World Records". I think it kind of falls apart there. He considers the following the best a horse can run, and thus equal on the speed scale:

Distance in furlongs, Time
4.0, 44.2
4.5, 50.4
5.0, 55.2
5.5, 61.4
6.0, 67.2
6.5, 73.6
7.0, 79.4
7.5, 86.6
8.0, 92.2
8.5, 98.4
9.0, 105.0
9.5, 112.4
10.0, 117.8

Why do I say it falls apart? Look at the difference adding an extra half furlong. For example, the first jump is 44.2 to 50.4, or 6.2. Here is the series:

6.2, 4.8, 6.2, 5.8, 6.4, 5.8, 7.2, 5.6, 6.2, 6.6, 7.4, 5.4

There is no way this can be an accurate baseline in my opinion. Horses would go from 9 to 9.5 furlongs in 7.4 seconds, then only take another 5.4 to run another 1/16 to make 10? Most of it doesn't make sense.

Any thoughts?

I think it is explained in the book that the "world records" are not linear and that there are jumps and bumps, but implies world record data can be used as a starting point - a wall that today's races are unlikely to cross but are theoretically possible to meet under ideal conditions.

If one record is slow then the gap to the next half furlong true fast distance is bound to be small. you cannot take the gaps as meaning anything except that the factual records are simply not linear. The 4/4.5 gap has now closed by races run since 1991 when the book was published. The 9/9.5 gap still stands from races run pre 1991.

The data is from all kinds of USA dirt etc tracks with possible timing errors and definitely true distance errors as well as different configurations and comes from horses of all ages and classes under different wind and weather conditions. The data does make perfect sense. Carrol shows how he has put a straight "best fit" line through the bumpy results to form a "world record estimation equation" with a correlation of accuracy of 0.99 - that removes the bumps if you want. Carrol shows how you "could" use actual track data with the "world data" to formulate your own ratings for that track in a consistent manner.

He does not ever claim 100% true accuracy and candidly explains the glitches - what more could you ask?

[Tom]

Have you tried all three points as a baseline and fit straights lines between the points for the various distances and let the chart self scale?

Mike (Dr Beav)

This might be a good idea for the other thread, about sprint and route par charts.

Most charts start with the $10K level and move up or down by a predetermined increment, but maybe it would better to use three baselines, high, low, average, and compare the two distributions so that Sprint High = Route High and so on, letting the data tell you what the scales are in reality.

[cj]

I found this pretty interesting:

http://www.chef-de-race.com/articles/speed_in_the_thoroughbred.htm

[pondman]

Carrol shows how you "could" use actual track data with the "world data" to formulate your own ratings for that track in a consistent manner.

If all horses at your track are slow and slower, you'll have too many problems trying to normalize data, if you are utilizing the world records. The point spread for $4,000 claimers will be small.

Although I'm not a fan of speed charts, I'd rather see someone establish their own chart for 4,000 claimers. But it's always going to be an average, with problems. Most of you already now when a horse is fast or super fast, and can toy with other horses. I'd stick with your intuition.

[PhantomOnTour]

Isn't the use of world records akin to using track records?
Aren't we going back into DRF speed rating-track variant stuff here?

Wouldn't a track record be more accurate than using a world record?

[cj]

Isn't the use of world records akin to using track records?
Aren't we going back into DRF speed rating-track variant stuff here?

Wouldn't a track record be more accurate than using a world record?

I would say no. It isn't perfect, but a world record is a world record. This isn't nearly as bad as making the best time at 6f at Beulah equal to the best time at 9f at Belmont.

[PhantomOnTour]

I would say no. It isn't perfect, but a world record is a world record. This isn't nearly as bad as making the best time at 6f at Beulah equal to the best time at 9f at Belmont.
Are we saying it's prudent to rate a 6f race at TuP and a 6f race at CRC against a world record?

Shippers would be a big problem with this method. If you go from CRC to TuP you will look like the slowest thing on the grounds using this method.

[cj]

Are we saying it's prudent to rate a 6f race at TuP and a 6f race at CRC against a world record?

Shippers would be a big problem with this method. If you go from CRC to TuP you will look like the slowest thing on the grounds using this method.

I didn't say it was perfect, I said it was better. You would still need to create variants track to track adjustments. At least though the baseline is somewhat more realistic. I certainly know it has flaws which is why I'm looking for better, much better.

[PhantomOnTour]

I make daily figs for Aqu-Aqui-Bel-Sar-GP-Mth as well as what I call Big Race Figs (BRF).
BRFs are graded stakes run on non turf at all major tracks in the USA...
CD-SA-Hol-Dmr-FG-Kee-Crc-Pim-Del-OP and some others i can't recall.
This requires a chart for each track or comparison to a single parallel time chart and adj made for each track.
I choose the former method; a unique chart for each track.
Why?
Because some tracks simply do not follow the orderly parallel time chart that says a horse who runs 1m in x-time will run 9f in y-time. It's much simpler for me to have a chart for each track rather than a load of adj. for each track.
It helps in track to track comparisons as shippers are in almost every race these days.
This method is clearly not perfect either, but i prefer it to a single chart or baseline.

[cj]

I make daily figs for Aqu-Aqui-Bel-Sar-GP-Mth as well as what I call Big Race Figs (BRF).
BRFs are graded stakes run on non turf at all major tracks in the USA...
CD-SA-Hol-Dmr-FG-Kee-Crc-Pim-Del-OP and some others i can't recall.
This requires a chart for each track or comparison to a single parallel time chart and adj made for each track.
I choose the former method; a unique chart for each track.
Why?
Because some tracks simply do not follow the orderly parallel time chart that says a horse who runs 1m in x-time will run 9f in y-time. It's much simpler for me to have a chart for each track rather than a load of adj. for each track.
It helps in track to track comparisons as shippers are in almost every race these days.
This method is clearly not perfect either, but i prefer it to a single chart or baseline.

This is pretty much exactly what I'm doing, but I want to start with a solid baseline. When there are deviations from track to track like you mention, I want to know why, and having that baseline helps a lot.

[PhantomOnTour]

This is pretty much exactly what I'm doing, but I want to start with a solid baseline. When there are deviations from track to track like you mention, I want to know why, and having that baseline helps a lot.
Is it time to construct a new universal parallel time chart, if an accurate one can even be created?
With AWS of many kinds in play today I think a universal baseline isn't doable.

[JeremyJet]

Is it time to construct a new universal parallel time chart, if an accurate one can even be created?
With AWS of many kinds in play today I think a universal baseline isn't doable.

There is no such thing as a "universal parallel time chart" ... one that can be used from track-to-track ... without adjustments ... never has been. Like you said, it's a "baseline" ... a starting point.

I don't know about Carroll, but Beyer never suggested otherwise.

[PhantomOnTour]

There is no such thing as a "universal parallel time chart" ... one that can be used from track-to-track ... without adjustments ... never has been. Like you said, it's a "baseline" ... a starting point.

I don't know about Carroll, but Beyer never suggested otherwise.
My error...i am not well versed on Beyer at all. I am a Quirin style pace and speed guy.
What I meant by universal chart is something that shows the standard (if you will) rate of deceleration from 8f to 8.5f to 9f to etc...horses getting 8f in 136.0 will get 9f in 149.8 or something like that.

[cj]

Is it time to construct a new universal parallel time chart, if an accurate one can even be created?
With AWS of many kinds in play today I think a universal baseline isn't doable.

I think you have to do it by surface, but the AWS are all pretty similar at least time wise.

[maddog42]

There is no such thing as a "universal parallel time chart" ... one that can be used from track-to-track ... without adjustments ... never has been. Like you said, it's a "baseline" ... a starting point.

I don't know about Carroll, but Beyer never suggested otherwise.

Lets say there is a mythical Worlds Greatest Handicapper. Please bear with me on this. He is handicapping AQU, BEL. WGH knows how to figure track variants like Einstein. HE can adjust every paceline to counter the effects of pace, wind ,heat and patrons passing gas. He knows where horses are in there form cycle. WGH KNOWS that on a day when the variants are equal that a 71.17 second 6f on AQUI is equal to 70.89 at belmont.WGH knows this because he can filter out all the other variables at these tracks and he has a Universal Parallel Time Chart that he constructed by being able to filter out all these other Factors that affect the outcome of a race and He still has a hard time overcoming the track takeout at SA.

[maddog42]

There is no such thing as a "universal parallel time chart" ... one that can be used from track-to-track ... without adjustments ... never has been. Like you said, it's a "baseline" ... a starting point.

I don't know about Carroll, but Beyer never suggested otherwise.

If you mean a Time Chart like what is in Picking Winners, then of course there is no such animal that will work on all tracks. I guess Univeral is the key word here.
I have constructed track charts that allowed me to go from one track to another fairly well. I would make an adjustment based on daily variant and let her rip. I have had more than a few people peering over my shoulder wanting to
know equivalents. That was a few years ago and keeping up with the changes was a nightmare. I did all this using notebooks, pencil and paper.

[JeremyJet]

Lets say there is a mythical Worlds Greatest Handicapper. Please bear with me on this. He is handicapping AQU, BEL. WGH knows how to figure track variants like Einstein. HE can adjust every paceline to counter the effects of pace, wind ,heat and patrons passing gas. He knows where horses are in there form cycle. WGH KNOWS that on a day when the variants are equal that a 71.17 second 6f on AQUI is equal to 70.89 at belmont.WGH knows this because he can filter out all the other variables at these tracks and he has a Universal Parallel Time Chart that he constructed by being able to filter out all these other Factors that affect the outcome of a race and He still has a hard time overcoming the track takeout at SA.

You're correct. Now think about the guy handicapping with inferior data. That guys really up shit creek.

[JeremyJet]

If you mean a Time Chart like what is in Picking Winners, then of course there is no such animal that will work on all tracks. I guess Univeral is the key word here.
I have constructed track charts that allowed me to go from one track to another fairly well. I would make an adjustment based on daily variant and let her rip. I have had more than a few people peering over my shoulder wanting to
know equivalents. That was a few years ago and keeping up with the changes was a nightmare. I did all this using notebooks, pencil and paper.

The purpose of universal charts is to cut down on the grunt work involved. Instead of starting from scratch for each track you want to make charts for, you can simply adjust the universal charts when you know what the distance-to-distance adjustments are for the particular track you're working with.

[BIG49010]

Just wondering if anybody uses an excellerated scale, being the faster races get more credit? I have thought about this in the past, but never put it to work.

I use an old set of Dave's pars, and have adjusted them over the years based on experience, to come up with my track to tracks.

[cj]

The purpose of universal charts is to cut down on the grunt work involved. Instead of starting from scratch for each track you want to make charts for, you can simply adjust the universal charts when you know what the distance-to-distance adjustments are for the particular track you're working with.

But Dr. Beav says this is absurd.

[SPEEDHORSE]

Bro CJ,

I am using Charles Carroll's Speed Figure methodology and I use a mythical track record as my baseline for all distances run here in the Philippines.

Since we only have two racetracks here, making a track to track adjustment is not problematic.

I calculate the elapsed time per distance per class and used the average elapsed time to add to my mythical track record. If my mythical track record for 1400 meters distance is 1:26.0. I will add to this baseline time the average elapsed time in seconds between 1400 and 1500 meters distance. If the average elapsed time is 6 seconds the baseline time for 1500 meters distance will be 1:26.0 and so on as the distances gets longer.

[highnote]

CJ -- check out Nick Mordin's book "On Time". It's about his method of making speed figures. He makes figures for British racing -- flat and jumps! Some of these tracks run downhill, some uphill, some on undulating courses, some clockwise and some counter clockwise. Champs Elysee has something like 17 different turf configurations!

His book might give you some good ideas on how to make adjustments.

His baseline time is the hypothetical grade one time at the track. Some tracks never have a grade one race, however, he can work it out based on shippers who have raced at both tracks.

[cj]

CJ -- check out Nick Mordin's book "On Time". It's about his method of making speed figures. He makes figures for British racing -- flat and jumps! Some of these tracks run downhill, some uphill, some on undulating courses, some clockwise and some counter clockwise. Champs Elysee has something like 17 different turf configurations!

His book might give you some good ideas on how to make adjustments.

His baseline time is the hypothetical grade one time at the track. Some tracks never have a grade one race, however, he can work it out based on shippers who have raced at both tracks.

I have his books, and I use his ideas often.

[highnote]

Check this out and see if it adds to the conversation.

http://www.chef-de-race.com/articles/speed_in_the_thoroughbred.htm


I think Steve Roman does really good work. It disappoints me that Beyer is such a critic of Dosage. Beyer's gripe is that Dosage is created after the fact. But the same can be said of Beyer speed figures. :D

But I digress.

At the link you gave, Roman writes about Dr. Fager running .40 faster than the projected world record time for a mile and this is the longest standing record.

Is it possible this race was mis-timed, or that the run-up was moved back? Shouldn't the time be measured from a standing start of exactly a mile?

I'd like to get the video and time it with a stopwatch or insert timecode into the video and try to figure out the actual time he ran for a mile.

A believe I read in Ragozin's book that they use timecode and measure from the starting gate.

[highnote]

I love this paragraph from the chef-de-race link:

The last kind of speed we need to discuss is that of instantaneous speed or, more realistically, fractional speed and pace. If one wishes to talk about the truly fastest horses, then there are a host of runners capable of getting the first quarter in :21 and a fraction or even less. But in actual fact, these types most often show their speed for only a brief time in the early stages of a race, usually tiring dramatically well before the finish. Often we call this "cheap speed". It is only cheap because it is not expressed over the full distance of a race. Actually it is not cheap at all. It is as valid a physiological expression as any other kind of speed. It just so happens that it occurs at one extreme of the speed frontier and well outside our normal frame of reference for meaningful racing. We view these runners in the same way we view runners that hold records at the other end of the spectrum, say 3 miles. In the end, we are drawn to speed displayed at distances that we, as a society, choose to race our horses. Within these bounds, all record holders are essentially equally fast.

[cj]

I think Steve Roman does really good work. It disappoints me that Beyer is such a critic of Dosage. Beyer's gripe is that Dosage is created after the fact. But the same can be said of Beyer speed figures. :D

But I digress.

At the link you gave, Roman writes about Dr. Fager running .40 faster than the projected world record time for a mile and this is the longest standing record.

Is it possible this race was mis-timed, or that the run-up was moved back? Shouldn't the time be measured from a standing start of exactly a mile?

I'd like to get the video and time it with a stopwatch or insert timecode into the video and try to figure out the actual time he ran for a mile.

A believe I read in Ragozin's book that they use timecode and measure from the starting gate.

I would love it if all races were timed from the gate, but I don't think we'll see it any time soon. I will say timing from the gate is tougher than some pretend, as is knowing the exact distance of gate to the pole.

Also, I apologize to "the judge" for posting the exact same link he did a few posts before mine. I missed that he had posted it.

[Valuist]

I think Steve Roman does really good work. It disappoints me that Beyer is such a critic of Dosage. Beyer's gripe is that Dosage is created after the fact. But the same can be said of Beyer speed figures. :D

But I digress.

At the link you gave, Roman writes about Dr. Fager running .40 faster than the projected world record time for a mile and this is the longest standing record.

Is it possible this race was mis-timed, or that the run-up was moved back? Shouldn't the time be measured from a standing start of exactly a mile?

I'd like to get the video and time it with a stopwatch or insert timecode into the video and try to figure out the actual time he ran for a mile.

A believe I read in Ragozin's book that they use timecode and measure from the starting gate.

You cannot possibly compare Beyer figures to dosage. Much of the alleged "Dosage streak" in the Derby was based on backfitting. The vast majority of Beyer numbers are not "back-fitted". And when they do, it is usually to acknowledge an initial mistake. But when you are dealing with one day of data, numerous distances, weather conditions and track maintenance, the possibility of error is far, far greater.

I seem to remember Creme Fraiche and Conquistador Cielo won the Belmont with insanely high dosage numbers. How did they get around it? They chalked it up to muddy track conditions. I think by now there's been enough non-qualifiers to win that is has been basically considered irrelevant.

[cj]

You cannot possibly compare Beyer figures to dosage. Much of the alleged "Dosage streak" in the Derby was based on backfitting. The vast majority of Beyer numbers are not "back-fitted". And when they do, it is usually to acknowledge an initial mistake. But when you are dealing with one day of data, numerous distances, weather conditions and track maintenance, the possibility of error is far, far greater.

I seem to remember Creme Fraiche and Conquistador Cielo won the Belmont with insanely high dosage numbers. How did they get around it? They chalked it up to muddy track conditions. I think by now there's been enough non-qualifiers to win that is has been basically considered irrelevant.

Personally, I think dosage has its place. I just don't think it was ever perfect and claiming it was able to predict the Derby and Belmont with perfection was silly. There is definitely a correlation between dosage and performance at various distances, but like anything else it is not infallible.

[highnote]

You cannot possibly compare Beyer figures to dosage. Much of the alleged "Dosage streak" in the Derby was based on backfitting. The vast majority of Beyer numbers are not "back-fitted". And when they do, it is usually to acknowledge an initial mistake. But when you are dealing with one day of data, numerous distances, weather conditions and track maintenance, the possibility of error is far, far greater.

I seem to remember Creme Fraiche and Conquistador Cielo won the Belmont with insanely high dosage numbers. How did they get around it? They chalked it up to muddy track conditions. I think by now there's been enough non-qualifiers to win that is has been basically considered irrelevant.


You are entitled to your beliefs, but I do not share them.

You are critical of Dosage. Fair enough. I have read and studied Roman's methods and find they are statistically sound, valid and relevant.

I use Beyer speed figures all the time.

Whether we're talking Beyers or Dosage they are both created using past data. That was my point, not whether they people find them useful or whether they are created the same way or can be compared as apples to apples. Rather than both being apples they are both fruit in that they both use past data.

Plenty of people are critical of Beyer and Dosage. But that doesn't change the facts of how they are made. Neither one attempts to predict the KY Derby, for example, but both can be useful in assessing how well a runner can compete in the Derby. A horse with high speed figures a pedigree skewed toward stamina should have a better chance in the Derby than a horse with low speed figures and a pedigree skewed toward sprints -- all else being equal.

However, slow horses with sprint pedigrees sometimes win route races. That's why we put them on the track. And Roman admits as much. He said the same thing with regards to Favorite Trick. He said its pedigree was tilted toward sprinting. His sire was Phone Trick -- a top quarterhorse sire. And if you ever saw Phone Trick on the breeding farm you would know just by looking at him that the horse would not stay beyond 6 furlongs. But that doesn't mean Favorite Trick can't stay longer. But statistically, Phone Trick produces sprinters. And if you're a breeder and you're breeding to Phone Trick you are probably not expecting your foal to win the KY Derby, but longshots sometimes happen. Maybe your foal will win. But you probably didn't pay the Phone Trick stud fee with the belief you had a strong chance of producing a KY Derby winner. However, in the auction ring you might have bought Favorite Trick and believed he would make a great race horse. You might have even thought he had the build of a horse that could stay 10 furlongs. But statistically Phone Trick is not going to produce many Favorite Tricks. And if he does start producing Favorite Tricks then his Dosage number will reflect that. Phone Trick will not become a Chef-De-Race until his offspring's racing sample size gets large enough that the type of race horse he produces can be statistically validated. I see nothing wrong with this method.

If you are trying to win the Derby you're going to breed to sires with Dosages like Alydar and Seattle Slew and in the auctions you're going to look at yearlings with low Dosage sires. You might even look at the Beyer figures earned by the low Dosage sires when they were racing.

However, when Strike the Gold won the Derby with a Dosage of something like 22, Alydar had not yet been classified as a Chef. But when enough of Alydar's runners started winning at classic distances Alydar was classified as a Chef.

There are horses today producing classic distance winners but they are not chefs. Producing one classic winner does not mean a sire will produce more. However, if a sire produces enough classic winners then you can say with some degree of confidence that the sire may continue to produce classic distance winners.

I could on. But this is enough. Probably no one is going to read this far. :D

[JeremyJet]

But Dr. Beav says this is absurd.

Absurd? That's a bit strong. I never knew Dr. Beav was an authority on speed figure methodology. Everyone is entitled to their opinions. I'll leave it at that.

[PhantomOnTour]

I would love it if all races were timed from the gate, but I don't think we'll see it any time soon. I will say timing from the gate is tougher than some pretend, as is knowing the exact distance of gate to the pole.

Also, I apologize to "the judge" for posting the exact same link he did a few posts before mine. I missed that he had posted it.
I remember a discussion awhile back about hand timing races from the gate. The start and finish would be no problem; neither would run up as that is published and adj. could be made. The problem was the fractions, as i recall, since it's tough to tell when the leader actually went by the designated pole, esp on the turns. Throw in some tall infield trees and the tote/big screen and you are bound to miss some poles on the backside.

[cj]

I remember a discussion awhile back about hand timing races from the gate. The start and finish would be no problem; neither would run up as that is published and adj. could be made. The problem was the fractions, as i recall, since it's tough to tell when the leader actually went by the designated pole, esp on the turns. Throw in some tall infield trees and the tote/big screen and you are bound to miss some poles on the backside.

Run up is published, but I'm not convinced it is all that accurately tracked by the chart guys. Final time only isn't too bad, but fractions, forget it. Even so, the human will not hand time as accurately as they think can. Try it from start to finish for the same race 5 times, you'll get 4 or 5 different readings and some of them are off by more than a hair.

[maddog42]

Even in that narrow range it looks fishy:

5.5-6.0: 5.8
6.0-6.5: 6.4
6.5-7.0: 5.8
Been meaning to reply to this post for a few days. Pick a track any track. The differences you come up with if you make a few pars (8 to 10) will vary widely.
5f to 5.5f might be 6.57
5.5f to 6f might be 6.26
6f to 6.5 might be 6.43
I have done this over many tracks and the progression is not linear. Why should we expect the world record to be linear? This is an indictment of the method and the main reason I started making PPF Pars. I am sure the runup is the main reason. If it varies, it makes the deceleration curve fall off faster or slower AT THE END OF THE RACE. I am making a major improvement (I hope) in the ppf Pars, but it involves 2nd call and energy distribution. HERESY! At least to Pizzolla and other pace purists. I have always enjoyed the role of heretic in politics and other things.
By the way if you want proof of the variation, I can point to many tracks with
wider variation than this.
Cj , I will let you do this. You may already be aware of this phenomenon. If so this comment will be for the less enlightened. This is the main reason I threw the 6.6 and 6.8 second adjustments of Pizzolla out the window in the first month of using it.

[maddog42]

I've read this book and like a lot of what he writes. I know others here do as well. However, the baseline he uses is based on "World Records". I think it kind of falls apart there. He considers the following the best a horse can run, and thus equal on the speed scale:

Distance in furlongs, Time
4.0, 44.2
4.5, 50.4
5.0, 55.2
5.5, 61.4
6.0, 67.2
6.5, 73.6
7.0, 79.4
7.5, 86.6
8.0, 92.2
8.5, 98.4
9.0, 105.0
9.5, 112.4
10.0, 117.8

Why do I say it falls apart? Look at the difference adding an extra half furlong. For example, the first jump is 44.2 to 50.4, or 6.2. Here is the series:

6.2, 4.8, 6.2, 5.8, 6.4, 5.8, 7.2, 5.6, 6.2, 6.6, 7.4, 5.4

There is no way this can be an accurate baseline in my opinion. Horses would go from 9 to 9.5 furlongs in 7.4 seconds, then only take another 5.4 to run another 1/16 to make 10? Most of it doesn't make sense.

Any thoughts?
Since you are critical of Beyers method in the other thread, I will take another shot at Carroll in this one. Carroll is lucky. He has a string of degrees and I am
sure I'm not the first to point this out. He goes to elaborate lengths to figure the length of a horse, even using a tape measure. He tries 8.5 and 9ft, but settles on 8, after trying even 7.5. He claims on his web site that the optimum length maybe 8.2 or 7.9 but not 8.5 or 7.5, because he tried them.
I believe him and I agree.
But. A big but.
He divides every race into 8ft lengths and computes his lengths based on the entire length of the race. He got lucky. 6f has 495 lengths in it.
Most of the time(some exceptions in turf racing) horses accelerate for the
first 2f. Then a slow deceleration follows. He decided to calculate the average
of the entire race to get his seconds per length. Usually this is .1400 to .1600
seconds per length. The acceleration at the beginning of the race couple with
the DECELERATION at the end saves him from the fire. I know this is nit picky,
but we are talking MONEY and accuracy here. He is usually off about 6 to 8%.
I know this because I compute the 3rd fraction about 90 times a racing day.

[classhandicapper]

I love this paragraph from the chef-de-race link:

I loved that paragraph because I recently started changing the way I think about horses.

Most people think of fast horses in terms of final time speed at the traditional distances, but to me fast is actually the maximum 2F-3F speed a horse can achieve (whenever it occurs within a race).

Stamina is how far the horse can sustain useful levels of speed.

They are independent qualities that each horse has in varying degrees and which is more significant is dependent on the conditions of the race (distance, surface, bias, pace, race development etc..)

IMO it can be useful to think about horses as individuals with different qualities than the more general way of thinking about their overall ability at "X" distance on "Y" surface. In some races I just want the fastest horse and in others I want the horse that can do battle and keep running.

[maddog42]

I make daily figs for Aqu-Aqui-Bel-Sar-GP-Mth as well as what I call Big Race Figs (BRF).
BRFs are graded stakes run on non turf at all major tracks in the USA...
CD-SA-Hol-Dmr-FG-Kee-Crc-Pim-Del-OP and some others i can't recall.
This requires a chart for each track or comparison to a single parallel time chart and adj made for each track.
I choose the former method; a unique chart for each track.
Why?
Because some tracks simply do not follow the orderly parallel time chart that says a horse who runs 1m in x-time will run 9f in y-time. It's much simpler for me to have a chart for each track rather than a load of adj. for each track.
It helps in track to track comparisons as shippers are in almost every race these days.
This method is clearly not perfect either, but i prefer it to a single chart or baseline.

Looking over the posts in this thread, I am assuming that most of you adjust your races %100. That is %100 of Track to Track adjustment ( YOUR PARS WILL TAKE CARE OF THAT) and %100 percent of daily variant. Have many of you experimented taking a percentage of daily variant? One of Carroll's contentions on his Website was that most handicappers overadjust there races. I have found this to be true. CJ has mentioned the small sample size throwing the variant off. This might be some some safeguard against this, always erring on the side of reality.

[cj]

Looking over the posts in this thread, I am assuming that most of you adjust your races %100. That is %100 of Track to Track adjustment ( YOUR PARS WILL TAKE CARE OF THAT) and %100 percent of daily variant. Have many of you experimented taking a percentage of daily variant? One of Carroll's contentions on his Website was that most handicappers overadjust there races. I have found this to be true. CJ has mentioned the small sample size throwing the variant off. This might be some some safeguard against this, always erring on the side of reality.

What I am developing is a way to increase the sample size. You look at all the horses that ran 1-2-3 and also any that finished within a certain number of lengths. You then project a range of variants for each race. Overlap the ranges and most times there are very few races that don't fit. I'm not saying this is easy, it is very complicated and I look at other factors that may have an impact on final time.

On many days, you have very little to go on. Even at a track with all dirt races, you might only have 8 races, and some of those will be horses with little if any history. The way I'm working on now uses at least 24 horses for an 8 race card, and often as many as 40 or more. It still isn't perfect, but it is better. Anything to increase the sample size is a good thing.

Here is an example from a recent day at Delmar:

8 Races:

Polytrack sprints, 2 races, one for cheap 2yo maiden fillies won easily by a 1st timer
Polytrack routes, 3 races, two for cheap maiden claimers
Turf Sprints, 1 race
Turf routes, 2 races, one for statebred maidens

One thing I have found is people tend to underestimate winners. Most winners improve. Another is people tend to overestimate beaten favorites.

[maddog42]

"One thing I have found is people tend to underestimate winners. Most winners improve. Another is people tend to overestimate beaten favorites".

If you mean most winners improve in their next race ? Depends. If you mean most horses improved off of previous race, to win I agree. Beaten Favorites ?
I am harsh on them.

[maddog42]

What I am developing is a way to increase the sample size. You look at all the horses that ran 1-2-3 and also any that finished within a certain number of lengths. You then project a range of variants for each race. Overlap the ranges and most times there are very few races that don't fit. I'm not saying this is easy, it is very complicated and I look at other factors that may have an impact on final time.

On many days, you have very little to go on. Even at a track with all dirt races, you might only have 8 races, and some of those will be horses with little if any history. The way I'm working on now uses at least 24 horses for an 8 race card, and often as many as 40 or more. It still isn't perfect, but it is better. Anything to increase the sample size is a good thing.

Here is an example from a recent day at Delmar:

8 Races:

Polytrack sprints, 2 races, one for cheap 2yo maiden fillies won easily by a 1st timer
Polytrack routes, 3 races, two for cheap maiden claimers
Turf Sprints, 1 race
Turf routes, 2 races, one for statebred maidens

One thing I have found is people tend to underestimate winners. Most winners improve. Another is people tend to overestimate beaten favorites.

While I find this more interesting than my question, You still didn't answer it.

This is a good idea, using not just the winner, but i feel (just a guess) that by the time you get down to the show horse, your data will be diluted by other factors such as pace, Jockey and horses just giving up. Was it Barry Meadow that said " living filth can finish 4th...". To a lesser extent 3rd also.
But Hey nothing succeeds like success, and if the data fits wear it. At least using the first 2 horses would double the sample size.

[cj]

While I find this more interesting than my question, You still didn't answer it.

This is a good idea, using not just the winner, but i feel (just a guess) that by the time you get down to the show horse, your data will be diluted by other factors such as pace, Jockey and horses just giving up. Was it Barry Meadow that said " living filth can finish 4th...". To a lesser extent 3rd also.
But Hey nothing succeeds like success, and if the data fits wear it. At least using the first 2 horses would double the sample size.

No, I don't use fractional parts of the variant.

Those factors you mention, at least pace, is included already.