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

Go Back   Horse Racing Forum - PaceAdvantage.Com - Horse Racing Message Board > Thoroughbred Horse Racing Discussion > Handicapper's Corner


Reply
 
Thread Tools Rating: Thread Rating: 3 votes, 5.00 average.
Old 02-21-2015, 09:20 PM   #1
osophy_junkie
Finish Line Profit
 
Join Date: May 2004
Posts: 143
Cool Projected Final Times for use in making Variants

I was recently remaking my Speed Figures and in the process of doing so I needed to create daily variants. I am currently just taking the final time minus the class par. And splitting this up by dirt/turf/sprint/route.

That got me to reading and doing some experimentation. Using simple track and distance conversion tables as a projection, I got the following results comparing to the current final time for the year of 2014. It is the sum of absolute difference between the two divided by the total number of races.

Comparing Class Par to Final Time.
Par (AQU) - 2.587
Par (SA) - 1.701

Comparing Previous Final time to Final Time.
Final Time (AQU) - 10.493
Final Time (SA) - 11.486

Projected Final time to Final Time.
Projected Final Time (AQU) - 4.0302
Projected Final Time (SQ) - 4.932

I tried a couple of other methods including Weighted Least Squares and SVM Regression, both showed worse results than just using the Previous Final Time. I plan on plugging in the Projected Time into my variant making function and measuring ROI next. What all do use to make projected figures? I'd really be interesting in hearing how you incorporate pace into the calculation.
osophy_junkie is offline   Reply With Quote Reply
Old 02-23-2015, 11:14 AM   #2
DeltaLover
Registered user
 
DeltaLover's Avatar
 
Join Date: Oct 2008
Location: FALIRIKON DELTA
Posts: 4,439
Quote:
Originally Posted by osophy_junkie
I was recently remaking my Speed Figures and in the process of doing so I needed to create daily variants. I am currently just taking the final time minus the class par. And splitting this up by dirt/turf/sprint/route.

That got me to reading and doing some experimentation. Using simple track and distance conversion tables as a projection, I got the following results comparing to the current final time for the year of 2014. It is the sum of absolute difference between the two divided by the total number of races.

Comparing Class Par to Final Time.
Par (AQU) - 2.587
Par (SA) - 1.701

Comparing Previous Final time to Final Time.
Final Time (AQU) - 10.493
Final Time (SA) - 11.486

Projected Final time to Final Time.
Projected Final Time (AQU) - 4.0302
Projected Final Time (SQ) - 4.932

I tried a couple of other methods including Weighted Least Squares and SVM Regression, both showed worse results than just using the Previous Final Time. I plan on plugging in the Projected Time into my variant making function and measuring ROI next. What all do use to make projected figures? I'd really be interesting in hearing how you incorporate pace into the calculation.

I am surprised that this posting did not attract any interest, getting zero responses although it refers to one of the most fundamental aspects of speed handicapping!

It seems that nobody has anything to say about the pragmatic and practical process of the creation of daily variants, speed figures and class pars but everyone is ready to jump in to theoretical discussions about if it is possible to show a consistent 20% ROI or whether Benter's multinomial logit modelling is applicable to North American Racing!!!!
__________________
whereof one cannot speak thereof one must be silent
Ludwig Wittgenstein
DeltaLover is offline   Reply With Quote Reply
Old 02-23-2015, 11:31 AM   #3
Capper Al
Registered User
 
Capper Al's Avatar
 
Join Date: Dec 2005
Location: MI
Posts: 6,330
I find that commerical speed figs (like BRIS) do a good enough job handling time and variants. I have put my extra effort into using feet per second approaches for an alternative views to speed figs. The commercial companies have deep pockets and resources to be able, in general, to produce a better speed fig than the lone handicapper can.
__________________


"The Law, in its majestic equality, forbids the rich, as well as the poor, to sleep under bridges, to beg in the streets, and to steal bread."

Anatole France


Capper Al is offline   Reply With Quote Reply
Old 02-23-2015, 11:57 AM   #4
sjk
Registered User
 
Join Date: Feb 2003
Posts: 2,105
Are you just projecting from the winner? I use any horse that runs top 3 or within 8 lengths of the winner and project from any top 3 or within 8 lengths
for races less than 90 days apart.
sjk is offline   Reply With Quote Reply
Old 02-23-2015, 12:34 PM   #5
raybo
EXCEL with SUPERFECTAS
 
raybo's Avatar
 
Join Date: Mar 2004
Posts: 10,206
To the OP, are you making "speed" figures or "performance" figures? "Speed" figures only measure final time. If your figures are going to include pace then they are not traditional "speed" figures.
__________________
Ray
Horseracing's like the stock market except you don't have to wait as long to go broke.

Excel Spreadsheet Handicapping Forum

Charter Member: Horseplayers Association of North America
raybo is offline   Reply With Quote Reply
Old 02-23-2015, 12:46 PM   #6
osophy_junkie
Finish Line Profit
 
Join Date: May 2004
Posts: 143
Quote:
Originally Posted by raybo
To the OP, are you making "speed" figures or "performance" figures? "Speed" figures only measure final time. If your figures are going to include pace then they are not traditional "speed" figures.
I was currently making speed figures, but would like to augment these to account for pace. So I guess I'll be making performance figures with this in mind.
osophy_junkie is offline   Reply With Quote Reply
Old 02-23-2015, 12:55 PM   #7
osophy_junkie
Finish Line Profit
 
Join Date: May 2004
Posts: 143
Quote:
Originally Posted by sjk
Are you just projecting from the winner? I use any horse that runs top 3 or within 8 lengths of the winner and project from any top 3 or within 8 lengths
for races less than 90 days apart.
I did some tests that I can't seem to find right now, but they didn't show much of an improvement when measuring the sum of entries that ranked in the top 3 and also had the highest 3 highest speed figures. I will try again and see how it effects ROI, really the only metric that matters.

So when using this to project times, do you then just average all 3 final times, to get what the race should look like?
osophy_junkie is offline   Reply With Quote Reply
Old 02-23-2015, 01:01 PM   #8
cj
@TimeformUSfigs
 
cj's Avatar
 
Join Date: Jan 2002
Location: Moore, OK
Posts: 46,816
My opinion, you need a LOT of experience to do projections these days. There are fewer races, more maiden races, more restricted claimers, and more turf races. And of the bigger tracks running turf races, there are multiple courses or "lanes" at several of them like Gulfstream, Belmont, Saratoga, Arlington, Laurel, etc. It isn't just that there are multiple lanes, but they use them on the same days.


Par times, in my opinion, just don't work. If I were going to start over today, rather than par times, I'd get a good set of figures for a few years going back and use them as projections rather than par times. Again, just my opinion.
cj is offline   Reply With Quote Reply
Old 02-23-2015, 01:40 PM   #9
sjk
Registered User
 
Join Date: Feb 2003
Posts: 2,105
I do all older non-statebred dirt races by track condition so for a typical day on a fast track there would be a dozen or two data points to average. I smooth out the far outliers.

I don't bet turf races so I have never used my turf numbers or bothered about obvious deficiencies noted.
sjk is offline   Reply With Quote Reply
Old 02-24-2015, 07:43 AM   #10
MJC922
Registered User
 
Join Date: Nov 2012
Posts: 1,506
Since I've had many years of speed figure-making, I'll pass along what my general process was before moving away from time altogether (for what it's worth).

It's good that you're seeking out wind measurements. Wind adjustments were step one in my process, break down all of the times for the day into segments and then adjust them for wind speed / direction. This step has no need for pars of any kind, what you need is global historical data which shows how much of an adjustment to make to the times of each segment when such and such wind speed hits a horse at a specific angle. So you need a function that accepts four parameters, wind speed, wind direction, horse's speed, horse's direction. For the latter you need to know what the avg angles are of each segment at that track.

So now moving on to step two as wind has been 'taken out of it' as Ragozin might say. Understand you can't do anything reliable with pace until wind has been taken out if it. Next step then, adjust for pace. This step requires pars but not CLASS pars, it needs fractional and final pars for that specific track / distance / surface. You need to flag the segments that are faster than par for each final time logged on the card and know what kind of impact that had on the final time. So you have another function to come up with, i.e. when SegmentA was this much faster than the par for that specific final time, how much to reduce the final time. Yes reduce it, it's always a reduction, 'less than ideal' pacing doesn't make anyone go faster, it only makes them go slower.

Again I would suggest as with the wind adjustments above use real historical data for this, never shortcut, no speculation, no guessing, not some theory of what it 'should be' if you're some physics expert. The problem is not going to be finding the impact of pace from the research but rather obtaining the fractional / final time averages at the track / distance / surface as you'll find many will have far too small of a sample size to be accurate within a fifth of a second let alone a tenth. If you don't know par pace for the final time then it's not really possible to effectively adjust for pace. in these cases you might as well leave pace out of the calculation. e.g. Try getting par pace for the Derby with however many 10f dirt races CD cards in a decade.

At this point, take your final times now and convert them into speed ratings. I'm not going to go over that part, most people are familiar with speed charts, Beyer, proportional time etc. You've brought quite a bit of value-added information into the calculations at this stage, you've taken both wind and pace 'out of it'.

Step three is the track variant, this is where you finally need the class pars (or the projections). I do like class pars in theory but in my opinion very few people have accurate ones. Unfortunately they really need to be accurate otherwise the track variant calculation (which depends upon a rather small sample size) will be skewed.

Anyway, (assuming you have class pars) at this point calculate the deviations from par. I wouldn't worry about a 'daily' variant, let's leave the daily part out of it... I would suggest taking the most recent race on the surface and work backward until you detect significant change, that'll be your track variant for the block of races. If the block goes back for a week before changing significantly, lucky you, if it goes back just two races these things happen too. Don't worry about breaking out by distance either, you already took wind out of it, so if your speed charts are good there should be no need to have separate variants by number of turns.

Ok, you made it to step four, here's where it gets ugly. Apply the variant to your speed rating and what numbers look to be way 'out of line' vs the class pars. This is where you build in a safety net which flags races for you (the human) to follow up on. Even before step one in the process make sure you haven't let any typos slip through. If you find it's a typo you can just fix and move on, if it's not a typo then you'll be forced to make a projection. Good luck to you in your time studies.

Last edited by MJC922; 02-24-2015 at 07:57 AM.
MJC922 is offline   Reply With Quote Reply
Old 02-24-2015, 11:35 AM   #11
osophy_junkie
Finish Line Profit
 
Join Date: May 2004
Posts: 143
Talking ROI data

And here is ROI data for betting the top figure for 2010-2014:

Just winners for variants.
Split just Winner(AQU) - 84.159%
Split just Winner(SA) - 84.936%
Split just Winner(GP) - 86.047%

Winner Projected.
Split w/ Projected(AQU) - 84.129%
Split w/ Projected(SA) - 79.725%
Split w/ Projected(GP) - 89.122%

Variants based on horses ITM
Split ITM(AQU) - 84.144%
Split ITM(SA) - 84.921%
Split ITM(GP) - 86.033%

Projects variants based on horses ITM
Split w/ Projected ITM(AQU) - 88.725%
Split w/ Projected ITM(SA) - 83.443%
Split w/ Projected ITM(GP) - 89.472%

Overall projecting the variant based on ITM entries, seems like the way to go. I wonder why SA is such an outlier.
osophy_junkie is offline   Reply With Quote Reply
Old 02-24-2015, 01:25 PM   #12
Capper Al
Registered User
 
Capper Al's Avatar
 
Join Date: Dec 2005
Location: MI
Posts: 6,330
Osophy_Junkie,

I would study fractional times using feet/sec ratios and compare the results to running style. Traynor's point to forget wind is good advice.
__________________


"The Law, in its majestic equality, forbids the rich, as well as the poor, to sleep under bridges, to beg in the streets, and to steal bread."

Anatole France


Capper Al is offline   Reply With Quote Reply
Old 02-24-2015, 02:56 PM   #13
cj
@TimeformUSfigs
 
cj's Avatar
 
Join Date: Jan 2002
Location: Moore, OK
Posts: 46,816
Here is my point on all of this...

If I had a machine that launched 1000 baseballs a day the same way every time, and on one day it averaged 400 feet, and on day two 380 feet, and day 3 420 feet, I could set the variants as 0, -20 feet, and +20 feet with no fear of being wrong.

Would it be possible to figure out why the days were different, and what factors contributed each day? Possibly. Would it help me get a better answer? Not really. Same goes for measuring track speed in my opinion.
cj is offline   Reply With Quote Reply
Old 02-24-2015, 03:27 PM   #14
classhandicapper
Registered User
 
classhandicapper's Avatar
 
Join Date: Mar 2005
Location: Queens, NY
Posts: 20,526
I like the idea of using Class PARs (where available) as a quality check against projections. I think it helps with quality control because the classing system is generally efficient. So if some figure is WAY out of line with the Class PAR, it's a red flag to at least take a better look at the race and make sure the figure makes sense. Granted, there are so many whacky classes it's hard to have a lot of good PARs, but the idea here is just to use them when available as a double check.

Also, Pars help prevent figure drift in case you have any small biases in your methodology or thinking that would cause your figures to slowly shrink or rise when they shouldn't be.

Some might argue that using PARs prevents you from drifting faster or slower if the horses are actually getting faster or slower over time, but I wouldn't worry much about that. With all the drugs (legal and legal), changes in care, changes in surfaces etc... it's close to impossible to compare horses from different time periods on figures anyway.

If I were to make my own figures again, it could only be for 1 track (maybe 2) because I'd incorporate Class Pars, run ups, some element of ground loss, my own bias notes, my own trip notes, and my own pace figures to make the projections. That's a huge job. It would have to be limited.

I think wind is a significant element to this, but I would only incorporate it when I had hard evidence of wind direction and force. Even then, I wouldn't use a formula. It would just be part of thinking if I also noticed that the fractional relationships looked way different than usual. I would then project the impact based on the results.
__________________
"Unlearning is the highest form of learning"

Last edited by classhandicapper; 02-24-2015 at 03:37 PM.
classhandicapper is online now   Reply With Quote Reply
Old 02-24-2015, 03:42 PM   #15
raybo
EXCEL with SUPERFECTAS
 
raybo's Avatar
 
Join Date: Mar 2004
Posts: 10,206
Quote:
Originally Posted by classhandicapper
I like the idea of using Class PARs (where available) as a quality check against projections. I think it helps with quality control because the classing system is generally efficient. So if some figure is WAY out of line with the PAR, it's a red flag to at least take a better look at the race and make sure the figure makes sense.

Also, Pars help prevent figure drift in case you have any small biases in your methodology or thinking that would cause your figures to slowly shrink or rise when they shouldn't be.

Some might argue that using PARs prevents you from drifting faster or slower if the horses are actually getting faster or slower over time, but I wouldn't worry much about that. With all the drugs (legal and legal), changes in care, changes in surfaces etc... it's close to impossible to compare horses from different time periods on figures anyway.
Wouldn't the same apply to class pars? "Class" levels/groupings, depending on how you determine them, and how tightly you group them, are also subject to changes in racing, from surface changes/maintenance and weather, to purse levels and other artificial manipulations including ever increasing sets of qualifications for races for carding different, supposed, degrees of "class" horses, to more and more frequent movements of horses between tracks, surfaces, distances, class levels, etc., the list goes on and on. Determining class levels is a minefield, full of obstacles that will, intuitively and calculatively (sp?), force significant misrepresentations of "class".
__________________
Ray
Horseracing's like the stock market except you don't have to wait as long to go broke.

Excel Spreadsheet Handicapping Forum

Charter Member: Horseplayers Association of North America
raybo 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 10:00 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.