Horse Racing Forum - PaceAdvantage.Com - Horse Racing Message Board

Go Back   Horse Racing Forum - PaceAdvantage.Com - Horse Racing Message Board > Thoroughbred Horse Racing Discussion > General Handicapping Discussion


Reply
 
Thread Tools Rate Thread
Old 10-30-2018, 07:19 PM   #61
classhandicapper
Registered User
 
classhandicapper's Avatar
 
Join Date: Mar 2005
Location: Queens, NY
Posts: 20,528
Quote:
Originally Posted by bobphilo View Post
Class,

I understand what you are saying about horses finishing hopelessly back and out of any purse money, but we are talking here about the difference between the top 2 finishers who are in contention for either 1st place or 2nd money. This also applies to 3 and 4 place finishers who are trying to get as much of the purse as possible.
I do agree that in many cases finish position may be of more importance than beaten lengths and you are correct in raising that issue. It is up to each handicapper to decide which measure is more important in each race situation.
This is very simple.

If you have some interest in the gaps between horses as a meaningful metric of any type (and the original poster must because he brought it up), it varies from position to position. So if you look at the gap between 1st and 2nd and try to apply that more broadly in some way, you will be making a lot of incorrect assumptions. I base that on the 10's of thousands of races I have already looked at for every distance, on every surface, on every track condition, and with every field size.

If he's only interested in winning margins, then he's got the information he wants, but I can't imagine why you would want to know winning margins unless you were working towards a more comprehensive picture of race results.

Beyond that, I was explaining some of the reasons why the data looks like it does.
__________________
"Unlearning is the highest form of learning"

Last edited by classhandicapper; 10-30-2018 at 07:23 PM.
classhandicapper is offline   Reply With Quote Reply
Old 10-30-2018, 07:56 PM   #62
steveb
Registered User
 
Join Date: Jan 2008
Location: Melbourne Australia
Posts: 900
Quote:
Originally Posted by classhandicapper View Post
This is very simple.

If you have some interest in the gaps between horses as a meaningful metric of any type (and the original poster must because he brought it up), it varies from position to position. So if you look at the gap between 1st and 2nd and try to apply that more broadly in some way, you will be making a lot of incorrect assumptions. I base that on the 10's of thousands of races I have already looked at for every distance, on every surface, on every track condition, and with every field size.

If he's only interested in winning margins, then he's got the information he wants, but I can't imagine why you would want to know winning margins unless you were working towards a more comprehensive picture of race results.

Beyond that, I was explaining some of the reasons why the data looks like it does.

i had not even thought of the gaps increasing, so thought i would check.
does not seem to support your opinion?


columns....
average speed point behind winner for # position
horse count in brackets and gap between it and preceding.[##].###


might try some different way to see what happens.

Attached Images
File Type: png HK_MORE.png (55.1 KB, 41 views)
steveb is offline   Reply With Quote Reply
Old 10-30-2018, 08:04 PM   #63
classhandicapper
Registered User
 
classhandicapper's Avatar
 
Join Date: Mar 2005
Location: Queens, NY
Posts: 20,528
Quote:
Originally Posted by steveb View Post
i had not even thought of the gaps increasing, so thought i would check.
does not seem to support your opinion?


columns....
average speed point behind winner for # position
horse count in brackets and gap between it and preceding.[##].###


might try some different way to see what happens.
The differences are not huge, but they are different. Some more than others, some not at all. I don't recall that there was some clear pattern to it when I broke it up. You have to adjust for field size because horses that finish way in the back of the field are sometimes eased or just terrible. That distorts the numbers.

To be clear, I am not looking at times. I am looking at lengths. I don't have all the horses times measured accurately at the finish.
__________________
"Unlearning is the highest form of learning"

Last edited by classhandicapper; 10-30-2018 at 08:13 PM.
classhandicapper is offline   Reply With Quote Reply
Old 10-30-2018, 08:22 PM   #64
classhandicapper
Registered User
 
classhandicapper's Avatar
 
Join Date: Mar 2005
Location: Queens, NY
Posts: 20,528
This is for dirt sprints.

They don't vary much at each sprint distance for each position. But you can see how they vary a bit between each position (you have to do the subtraction manually). They are large samples. The smallest one is for 5F and even that sample is about 500 races. The rest are all above 1000 and some greater than 5000.

Attached Images
File Type: png sprint.PNG (16.1 KB, 29 views)
__________________
"Unlearning is the highest form of learning"

Last edited by classhandicapper; 10-30-2018 at 08:31 PM.
classhandicapper is offline   Reply With Quote Reply
Old 10-30-2018, 08:32 PM   #65
steveb
Registered User
 
Join Date: Jan 2008
Location: Melbourne Australia
Posts: 900
Quote:
Originally Posted by classhandicapper View Post
The differences are not huge, but they are different. Some more than others, some not at all. I don't recall that there was some clear pattern to it when I broke it up. You have to adjust for field size because horses that finish way in the back of the field are sometimes eased or just terrible. That distorts the numbers.

To be clear, I am not looking at times. I am looking at lengths. I don't have all the horses times measured accurately at the finish.

there was at least 8 starters in every race.
the only way you can confidently account for field size if you don't have enough is to standardise or normalise them.


your last sentence does not make sense because lengths come from time, so there would be very strong correlation between the two.
as an indication of that correl, here is something i do every time i add data as a check.
i make 2 files every meeting, here is one that does each race correlation between margin and time.
this enables me to pick up errors....not that it ever happens in hk but it certainly does most other joints.


date/track/racetype/correlation between time&margin/conversion factor
conversion factor is always .16 but because the margin are in increments of .25 then never perfect, but it's the margin that's wrong not the time.
Attached Files
File Type: xlsx CONVERSION_FACTOR_CHECKER.xlsx (27.8 KB, 44 views)
steveb is offline   Reply With Quote Reply
Old 10-30-2018, 08:47 PM   #66
classhandicapper
Registered User
 
classhandicapper's Avatar
 
Join Date: Mar 2005
Location: Queens, NY
Posts: 20,528
Quote:
Originally Posted by steveb View Post
there was at least 8 starters in every race.
the only way you can confidently account for field size if you don't have enough is to standardise or normalise them.


your last sentence does not make sense because lengths come from time, so there would be very strong correlation between the two.
as an indication of that correl, here is something i do every time i add data as a check.
i make 2 files every meeting, here is one that does each race correlation between margin and time.
this enables me to pick up errors....not that it ever happens in hk but it certainly does most other joints.


date/track/racetype/correlation between time&margin/conversion factor
conversion factor is always .16 but because the margin are in increments of .25 then never perfect, but it's the margin that's wrong not the time.
The reason I mentioned I was using lengths is that I don't know the exact conversion formula for beaten lengths to time in the US.

In the data I just posted I only displayed finish positions 2-4. That way field size is not much of an issue. If there is any distortion because a handful of fields only had 3-4 horses and others had more I doubt it had a huge impact. It's the horses way in the back of the field that distort beaten lengths averages because they often get eased.

When I studied the data initially, I broke each set out by the field size.

It was by Distance, Surface, Track Condition, Field Size, Finish Position, Beaten lengths for every combination.

I started this project trying to improve some ratings I create for myself. I was looking for a relatively simple formula I could use. The more I looked at the data the more I realized there was no simple solution to what I wanted. So I dropped it. I focused primarily on finish position with field size and beaten lengths being secondary. That worked better. I'm sure I could make the system much better, but it's good enough for my purposes. I always dig into races in a more detailed way than I can program anyway.
__________________
"Unlearning is the highest form of learning"

Last edited by classhandicapper; 10-30-2018 at 08:49 PM.
classhandicapper is offline   Reply With Quote Reply
Old 10-30-2018, 09:30 PM   #67
steveb
Registered User
 
Join Date: Jan 2008
Location: Melbourne Australia
Posts: 900
here is another one that might give somebody an idea as far as error checking routines go.
all done programatically.
if something looks weird the computer automatically tells you it thinks something is up that creek.



first 4 columns self explanatory
M_TIME is converting the margin back to a time to see if it matches the official time.....never perfect because the margins are not perfect being .25 increments
MARGIN is official margin
T_MARGIN is converting time to margin to make sure it fits as well as possible
CON_FACTOR is trying to fiigure the conversion factor for each horse.
and again it can't be perect because of those increments, and especially when margin is small. should be .16 in hk
Attached Files
File Type: xlsx Check_Whole_Meeting_MargTime_Conversions_HPVY.xlsx (18.0 KB, 30 views)
steveb is offline   Reply With Quote Reply
Old 10-30-2018, 09:54 PM   #68
steveb
Registered User
 
Join Date: Jan 2008
Location: Melbourne Australia
Posts: 900
Quote:
Originally Posted by steveb View Post
here is another one that might give somebody an idea as far as error checking routines go.
all done programatically.
if something looks weird the computer automatically tells you it thinks something is up that creek.



first 4 columns self explanatory
M_TIME is converting the margin back to a time to see if it matches the official time.....never perfect because the margins are not perfect being .25 increments
MARGIN is official margin
T_MARGIN is converting time to margin to make sure it fits as well as possible
CON_FACTOR is trying to fiigure the conversion factor for each horse.
and again it can't be perect because of those increments, and especially when margin is small. should be .16 in hk

and i should mention...
if TIME lesss M_TIME is negative
then MARGIN less T_MARGIN must be positive
and the inverse is true

if one is zero then the other must be zero too


if those conditions are not met then you know something is wrong.
steveb is offline   Reply With Quote Reply
Old 11-02-2018, 05:29 PM   #69
bobphilo
Registered User
 
Join Date: Dec 2005
Location: Palm Beach, Florida
Posts: 2,465
Quote:
Originally Posted by steveb View Post
there was at least 8 starters in every race.
the only way you can confidently account for field size if you don't have enough is to standardise or normalise them.


your last sentence does not make sense because lengths come from time, so there would be very strong correlation between the two.
as an indication of that correl, here is something i do every time i add data as a check.
i make 2 files every meeting, here is one that does each race correlation between margin and time.
this enables me to pick up errors....not that it ever happens in hk but it certainly does most other joints.


date/track/racetype/correlation between time&margin/conversion factor
conversion factor is always .16 but because the margin are in increments of .25 then never perfect, but it's the margin that's wrong not the time.
When I was making my own speed and pace figures I would convert beaten lengths to time and rate the horses on that. I also found that the conversion factor of .16 worked very well for U.S. races too. Of course the this can vary somewhat depending on the speed of the horses at that point of call and a more precise time could be arrived at by using the horse's velocity at that sectional. Not sure if it's worth the extra effort but it can be done.
Interesting that you found the same conversion factor in Hong Kong racing but not surprising.
bobphilo is offline   Reply With Quote Reply
Old 11-03-2018, 11:34 AM   #70
DeltaLover
Registered user
 
DeltaLover's Avatar
 
Join Date: Oct 2008
Location: FALIRIKON DELTA
Posts: 4,439
This thread presents enough data to refute Bayer’s conjecture about the diminished significance of beaten lengths in proportion to distance.

Still, even if we can reject one of the most fundamental axioms of Beyer’s theory, we cannot compare it against other approaches (that potentially can contain equally erroneous assumptions) unless we develop some comparison mechanism that will be able to operate directly of data (maintaining a complete isolation from the algorithmic details is that are used in the implementation layer).

This mechanism should receive a set of speed figures and the corresponding race results and produce a sorted list of them using criteria such as predictability or profit making it possible to decide about the relative effectiveness of each figure making methodology.

Any ideas how such a mechanism can be created?
__________________
whereof one cannot speak thereof one must be silent
Ludwig Wittgenstein
DeltaLover is offline   Reply With Quote Reply
Old 11-03-2018, 11:47 AM   #71
cj
@TimeformUSfigs
 
cj's Avatar
 
Join Date: Jan 2002
Location: Moore, OK
Posts: 46,816
Quote:
Originally Posted by DeltaLover View Post
This thread presents enough data to refute Bayer’s conjecture about the diminished significance of beaten lengths in proportion to distance.

Still, even if we can reject one of the most fundamental axioms of Beyer’s theory, we cannot compare it against other approaches (that potentially can contain equally erroneous assumptions) unless we develop some comparison mechanism that will be able to operate directly of data (maintaining a complete isolation from the algorithmic details is that are used in the implementation layer).

This mechanism should receive a set of speed figures and the corresponding race results and produce a sorted list of them using criteria such as predictability or profit making it possible to decide about the relative effectiveness of each figure making methodology.

Any ideas how such a mechanism can be created?
To be fair, Beyer is merely using beaten lengths as a way to calculate the time of the also rans. Every horse is rated on the time it ran. I think this thread shows there could be some evolving to a better way, that is all.
cj is offline   Reply With Quote Reply
Old 11-03-2018, 12:09 PM   #72
Tom
The Voice of Reason!
 
Tom's Avatar
 
Join Date: Mar 2001
Location: Canandaigua, New york
Posts: 112,470
Bottom line, now much difference in the final fig sis it going to make?

If we had better fgis, would we have seen Marley's Freedom would be .75 lengths short of winning the FM Sprint?
__________________
Who does the Racing Form Detective like in this one?
Tom is offline   Reply With Quote Reply
Old 11-03-2018, 12:20 PM   #73
classhandicapper
Registered User
 
classhandicapper's Avatar
 
Join Date: Mar 2005
Location: Queens, NY
Posts: 20,528
Quote:
Originally Posted by DeltaLover View Post
This thread presents enough data to refute Bayer’s conjecture about the diminished significance of beaten lengths in proportion to distance.
If you want my honest opinion, I think the speed figure model of handicapping is simply wrong assuming your goal is to understand what's actually going on at the track.

It's more about the quality of the field, how the race develops, and how the track is playing. The fractions and final times reflect the interaction between those 3 things, but they don't necessarily reflect how well the horses ran even if you had perfectly accurate pace and final time numbers (which you often won't). How competitively the race is run, the horse's positions, the moves, how tiring the track is etc.. matter a lot and can cause wide variations in numbers.

The reason speed figures pick so many winners is that better horses tend to produce a "faster outcomes" and horses that finish near the front are typically the better horses. So better horses tend to have faster figures.

But IMHO, if you want to create the "correct" model of evaluating horses it has more to do with an analysis of the field going into a race (how fast they generally are, what their running styles are, what quality of horses they've been running against), watching the race develop, and analyzing the result chart and race flow to determine how that track and development impacted the outcome.

You can create metrics that measure these things that can be used as tools in the analysis (including pace and speed figures), but in the end it's better to subjectively analyze a race using those tools and NOT let the numbers and theories dictate to you what you think happened. The results on the track speak for themselves and sometimes they reflect reality better than the numbers. It's a matter of getting better at understanding it all in a comprehensive way. No easy task.
__________________
"Unlearning is the highest form of learning"

Last edited by classhandicapper; 11-03-2018 at 12:34 PM.
classhandicapper is offline   Reply With Quote Reply
Old 11-03-2018, 12:58 PM   #74
DeltaLover
Registered user
 
DeltaLover's Avatar
 
Join Date: Oct 2008
Location: FALIRIKON DELTA
Posts: 4,439
Quote:
Originally Posted by classhandicapper View Post
If you want my honest opinion, I think the speed figure model of handicapping is simply wrong assuming your goal is to understand what's actually going on at the track.

It's more about the quality of the field, how the race develops, and how the track is playing. The fractions and final times reflect the interaction between those 3 things, but they don't necessarily reflect how well the horses ran even if you had perfectly accurate pace and final time numbers (which you often won't). How competitively the race is run, the horse's positions, the moves, how tiring the track is etc.. matter a lot and can cause wide variations in numbers.

The reason speed figures pick so many winners is that better horses tend to produce a "faster outcomes" and horses that finish near the front are typically the better horses. So better horses tend to have faster figures.

But IMHO, if you want to create the "correct" model of evaluating horses it has more to do with an analysis of the field going into a race (how fast they generally are, what their running styles are, what quality of horses they've been running against), watching the race develop, and analyzing the result chart and race flow to determine how that track and development impacted the outcome.

You can create metrics that measure these things that can be used as tools in the analysis (including pace and speed figures), but in the end it's better to subjectively analyze a race using those tools and NOT let the numbers and theories dictate to you what you think happened. The results on the track speak for themselves and sometimes they reflect reality better than the numbers. It's a matter of getting better at understanding it all in a comprehensive way. No easy task.
I am not sure what you mean here.

Your goal of “understanding what is going on at the track” cannot be empirically validated as it is open to multiple interpretations thus it cannot be the objective a statistical model.

Detecting the “wrongness” assumes that you have a definition for “correctness” and this is where the real challenge is hidden.

The input to a speed figure model consists of measurable quantities that generate deterministic output; what I am saying here is that while the validation of this “output” is the most critical part of the whole process I am not aware of any related methodology.
__________________
whereof one cannot speak thereof one must be silent
Ludwig Wittgenstein

Last edited by DeltaLover; 11-03-2018 at 01:01 PM.
DeltaLover is offline   Reply With Quote Reply
Old 11-03-2018, 01:00 PM   #75
DeltaLover
Registered user
 
DeltaLover's Avatar
 
Join Date: Oct 2008
Location: FALIRIKON DELTA
Posts: 4,439
Quote:
Originally Posted by cj View Post
To be fair, Beyer is merely using beaten lengths as a way to calculate the time of the also rans. Every horse is rated on the time it ran. I think this thread shows there could be some evolving to a better way, that is all.
To be more precise he is using beaten lengths in conjunction to the distance of the race; this is the assumption that is invalidated by the data presented here.
__________________
whereof one cannot speak thereof one must be silent
Ludwig Wittgenstein
DeltaLover is offline   Reply With Quote Reply
Reply




Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

» Advertisement
Powered by vBadvanced CMPS v3.2.3

All times are GMT -4. The time now is 02:36 AM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2024, vBulletin Solutions, Inc.
Copyright 1999 - 2023 -- PaceAdvantage.Com -- All Rights Reserved
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program
designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.